Contents Introduction
In the near future, the world is set to experience a significant increase in global mobile data traffic. Predictions indicate that by 2030, this data traffic could reach 5016 exabytes per month (EB/mo), marking significant growth from the 7.462 EB/mo recorded in 2010 [1]. This increasing demand is primarily driven by the emergence of new sixth-generation (6G) wireless applications, such as augmented reality [2], virtual reality [3], autonomous vehicles [4], digital health [5], and smart cities [6]. These cutting-edge technologies require significant improvements in system performance, including ultra-high data rates, ultra-low latency, and highly efficient wireless connectivity. To address these growing demands and improvements, a paradigm shift from evolutionary progress in wireless networks to a revolutionary approach is required. The focus should move beyond simply improving base stations (BSs) and intelligent devices, towards a revolutionary approach that enables intelligent reconfiguration of the propagation environment [7]. Consequently, numerous innovative approaches and technologies have been proposed in the literature, aiming to seamlessly integrate with 6G systems and provide the capabilities needed to meet the ever-increasing demands of modern communication requirements. Among these solutions, the concept of controlling propagation via reconfigurable intelligent surfaces (RIS) has received significant attention [8].
RIS can reconfigure the wireless propagation environment by utilizing a large number of low-cost passive meta-surface reflection units (RUs) capable of intelligently adjusting the incident signal to improve system performance [9]. In particular, these surfaces can effectively mitigate interference and reduce noise by intelligently manipulating the phase shifts of the reflected signals and dynamically adjusting the signal paths. This results in cleaner and more reliable received signals. Additionally, RIS optimizes the signal paths by reflecting and focusing the signals directly toward the receiver, combating signal degradation and improving signal strength and quality [10]. By leveraging RIS, signals can be transmitted from a source node to a destination node by reflecting radio waves. This innovative approach allows the destination node to receive signals that combine both direct line-of-sight (LoS) and reflected links. Even in scenarios where only a non-line-of-sight (NLoS) link is available, the destination node can still receive signals through the reflected path, as shown in Figure 1.
A. Horizons of RIS Applications
Breaking away from conventional cooperative communication systems with relay technology, an RIS gains control over communications by employing an artificial planar surface [11]. The challenge of power allocation among wireless users in non-orthogonal multiple-access (NOMA) networks was effectively tackled by RIS in [12]. Likewise, the authors in [13], demonstrated the effectiveness of utilizing RIS as a relay to enhance the throughput performance of backscatter link systems among wireless users. The capacity optimization was explored in [14], wherein RIS was utilized and the coefficients for phase shift as well as the transmit covariance matrices were optimized, enabling efficient signal distribution among multiple transmitters and receivers. Beyond terrestrial applications, RIS can also be deployed in the air, often in conjunction with unmanned aerial vehicles (UAVs). For example, the utilization of an aerial RIS was suggested in [15] to optimize power distribution and enhance network coverage, specifically in microwave links. Similarly, in [16] a joint optimization of active and passive beamforming at the BS and RIS that maximizes the data rate has been carried out in integrated aerial RIS-assisted NOMA networks. In addition to being transported through the air by UAVs, RISs can also be affixed to walls within a room or building. This was exemplified in [17], where an RIS was strategically positioned on a wall to enable indoor multi-user localization. Likewise, in [18], [19], and [20], an RIS was deployed on a building to facilitate communication between a mobile UAV, a fixed position UAV, and ground users. Additionally, the versatility of RIS technology extends to underground or underwater applications, as evidenced in [21], [22], and [23]. In particular, the authors in [21] utilized RIS to enhance the signal-to-noise ratio (SNR) performance in underground parking lots. Furthermore, the authors in [22] and [23] employed RIS to improve outage performance and boost channel capacity in mixed communication of radio frequency (RF) and underwater wireless optical communication (UWOC) systems. Moreover, RIS exhibited the capability to serve as a central point to support massive signal reflection in device-to-device (D2D) communication networks [24]. In addition, RIS effectively compensated for power losses that occurred over extended distances in networks designed for simultaneous wireless information and power transfer (SWIPT) [25]. Furthermore, RIS has found applications in the Internet of Things (IoT) networks, contributing to intelligent wireless sensor networks, precision agriculture, and advanced industrial facilities [26].
Although the focus of this review paper is on RF-based wireless communication, RIS also has shown promise in non-RF wireless technologies. For instance, RIS can be used to establish virtual LoS links in visible light communication (VLC) networks [27], mitigate signal attenuation in free space optics (FSO) systems [28], and optimize the integration of RF and VLC/FSO technologies in hybrid systems [29], [30]. Figure 2 provides a visual representation of the wide range of applications where RISs can be utilized as passive components [31]. This highlights the potential of RIS technology to transform communication systems in diverse environments.
![FIGURE 2. - Illustration of RIS applications adapted from [11].](/mediastore/IEEE/content/media/6287639/10005208/10356072/mahmu2-3342320-large.gif)
Despite the innovative opportunities that RIS introduces for enhancing wireless communication, it also raises several potential security implications and vulnerabilities.
For example, one potential security concern is the risk of unauthorized access or manipulation of the RIS control mechanisms [32]. Malicious actors could attempt to interfere with the phase and amplitude adjustments of RIS elements, potentially disrupting signal paths and compromising network integrity. Additionally, the use of RIS for eavesdropping, where RIS elements are manipulated to redirect signals to unintended recipients, poses a privacy threat [33]. To address these security vulnerabilities, various mitigation strategies have been proposed in the literature [34]. These strategies include implementing robust authentication and encrypted access control mechanisms to ensure that only authorized entities can adjust RIS configurations. Furthermore, the integration of artificial intelligence (AI) and machine learning (ML) algorithms can enhance network security by identifying and preventing suspicious activities in real-time. Finally, proper encryption of signals and communications can protect against eavesdropping, ensuring data privacy. More in-depth details of security implications or vulnerabilities associated with RIS-aided networks and how they can be mitigated can be found in [34].
B. Potential of RIS for 6G Wireless Systems
The introduction of RIS in 6G systems represents a significant departure from the evolutionary trends seen in previous generations of wireless communication systems [35]. Previous generations of wireless communication systems focused on enhancing the capabilities of wireless networks primarily through advancements in hardware and spectral efficiency (SE). For instance, fourth-generation (4G) wireless communication systems introduced significant improvements in data rates and latency compared to third-generation (3G) ones, mainly by optimizing the use of available spectrum and deploying more advanced modulation techniques [36]. Meanwhile, the fifth-generation (5G) marked a notable advancement as it introduced technologies like massive multiple input multiple outputs (mMIMO), millimeter wave (mmWave) communication, and network slicing [37], [38]. These innovations significantly improved data rates, network capacity, and SE while reducing latency, setting it apart from 4G. However, 6G represents a paradigm shift. While it continues the trend of improving data rates, latency, and spectral efficiency, it goes beyond these conventional metrics. 6G recognizes that simply optimizing the existing wireless infrastructure will not be enough to meet the needs of the diverse and demanding applications of the future.
RIS is a new technology that introduces an entirely new dimension to wireless communication by intelligently controlling signal reflections. Unlike previous advancements that primarily relied on centralized, power-hungry BSs, RIS uses passive surfaces to manipulate electromagnetic waves [9]. These surfaces offer a range of advantages, including cost-effectiveness, and the absence of a major power supply, complex signal processing, or encoding and decoding processes [39]. Moreover, RIS can be integrated into various environments, from urban areas to factories and even inside vehicles. This makes it highly adaptable and flexible, which is essential for supporting the diverse and demanding applications of future 6G wireless systems. This newness is similar to the transition from traditional cellular networks to IoT networks. Just as IoT devices extended the concept of connectivity beyond smartphones and computers to everyday objects, RIS extends the capabilities of wireless networks beyond conventional BSs and antennas to intelligent surfaces. Furthermore, RIS is also part of a broader ecosystem that includes AI [40], terahertz (THz) communication [41], and VLC system, creating a partnership that is essential for the 6G vision. These technologies are working together to create a new era of wireless communication that will be ultra-reliable, low-latency communication, which is capable of supporting a wide range of new and innovative applications Importantly, integrating RIS into 6G systems has positive energy consumption implications. RIS devices are inherently passive, requiring minimal power to perform signal manipulation, aligning with the energy-efficient goals of 6G. By optimizing signal paths, reducing interference, and enabling lower transmit power, RIS enhances overall energy efficiency (EE) and contributes to greener and more sustainable networks. This aligns with the broader trend of “green” or environmentally sustainable communication systems. This emphasis on sustainability is more pronounced compared to earlier generations.
However, the widespread adoption of RIS in real-world 6G deployments faces several primary technical barriers. Firstly, the complex optimization required for RIS configurations in dynamic wireless environments, particularly in large-scale systems is still elusive. This optimization can be computationally intensive and time-consuming, necessitating the development of efficient algorithms and strategies to tackle this complexity. Moreover, accurate channel state estimations are essential for optimizing RIS configurations effectively and intelligently reflecting signals [42]. Inaccurate channel estimation can lead to suboptimal phase adjustments, reducing the ability of the RIS to optimize signal strength and therefore impacting network performance. However, estimating the channel state information (CSI) in real-time, especially in complex environments with high mobility, can be challenging. Noise, interference, and delays can degrade the quality of CSI, making techniques for efficient and reliable channel state estimation that guarantee robustness against these factors compulsory. Furthermore, the practical implementation of RIS relies on cost-effective, energy-efficient, and compact hardware elements that can operate at high frequencies, which is particularly challenging for mmWave and THz bands. Additionally, discrete phase shifts required for practical RIS implementation introduce quantization errors. These errors can lead to suboptimal performance, especially in scenarios requiring accurate phase control. The impact of quantization depends on factors such as bit resolution; higher bit resolution allows for accurate control of phase shifts but requires more feedback information and may lead to increased power consumption. Conversely, lower resolution simplifies feedback but can limit the achievable performance. Balancing the trade-offs between quantization resolution, feedback overhead, and quantization error, is complex but essential for optimizing signal quality and achieving efficient RIS deployment. Achieving scalable solutions for large areas, seamless integration with existing networks, and demonstrating practical feasibility in diverse deployment scenarios are also substantial technical barriers confronting the uptake of RIS-aided networks. Finally, managing interference in scenarios with multiple RIS devices and users, and ensuring security against eavesdropping and unauthorized configuration changes, further add to the complexity of RIS adoption in 6G networks.
Although there is substantial theoretical research supporting the potential of RIS to enhance signal strength and quality, specific case studies and real-world applications demonstrating significant improvements are still emerging. Some preliminary real-world experiments and simulations have been conducted and demonstrated significant improvements in signal strength and quality. For instance, the authors in [43] have developed an RIS prototype with 256 RUs, each with 2-bit programmable phase shifts. This RIS prototype was tested in an anechoic chamber at a frequency of 2.3 GHz, achieving an antenna gain of 21.7 dBi. This result suggests that RIS has the potential to significantly enhance the signal strength and coverage in wireless communication systems, particularly at the specified frequency. Moreover, the authors in [44] conducted a simulation to assess the potential of RIS in a scenario where signal strength is severely compromised due to signal blockages. The results demonstrated that activating the RIS led to an improvement in received signal strength by more than 15 dB. This indicates that RIS has the potential to effectively mitigate signal blockages and significantly boost signal quality. However, there have been limited investigations into the effectiveness of RIS in large-scale, widely recognized operational commercial networks so far. In [45] and [46], the researchers conducted real-world experiments in 5G commercial networks to evaluate whether RIS enhances coverage and signal quality. The practical findings confirmed that RISs are indeed effective in addressing coverage challenges and improving signal quality. Recently, a comprehensive evaluation of RIS systems operating at both 5.8 GHz and 2.6 GHz was conducted in [47]. The evaluation covered various scenarios, including office and corridor setups, as well as outdoor tests involving operational 5G networks. The results showed that in office and corridor setups, the 5.8 GHz RIS demonstrated a significant power gain of 10–20 dB at the receiver. This improvement in power gain was even more impressive in outdoor tests, where the 5.8 GHz RIS offered a remarkable power gain of 35 dB. These findings highlight the potential of RIS systems to enhance wireless communication in different environments. For the 2.6 GHz RIS within commercial 5G networks, the evaluation showed that RIS could enhance indoor coverage by 4–7 dB. This improvement in coverage is particularly promising for improving connectivity in indoor spaces, where signal strength can often be a challenge. Furthermore, the study explored these experiments considering both directional and omnidirectional antennas. The results were quite enlightening. It was found that equipping transceivers with omnidirectional antennas resulted in insufficient gains, with only a 4 dB improvement. On the other hand, directional antennas offered a substantial 15 dB improvement. This performance disparity between directional and omnidirectional antennas presents a barrier that needs to be addressed for the widespread adoption of RIS systems. This challenge arises from the fact that RIS technology works best when it can precisely steer signals in specific directions. Omnidirectional antennas, which radiate signals in all directions, do not align well with the specific requirements of RIS. Consequently, for RIS to be effectively commercialized, there is a need for compatibility between the type of antennas used in transceivers and the capabilities of RIS. Using omnidirectional antennas in combination with RIS may not fully exploit the full potential of the technology. Conversely, integrating directional antennas can introduce complexities and higher costs to the infrastructure, as these antennas are more specialized and expensive compared to their omnidirectional counterparts. This cost factor could pose a significant barrier to the widespread adoption of RIS technology. Therefore, these considerations highlight the need for further research to overcome this limitation, with more careful planning, and innovative solutions to facilitate the seamless integration and cost-effective deployment of RIS technology in practical commercial applications.
C. Optimizing Reflection Properties: A Paradigm Shift in Wireless Communication
The reflection properties controlled by RIS differ significantly from traditional wireless communication methods [48]. In conventional wireless communication, signals propagate directly from the transmitter to the receiver through free space often encountering obstacles, interference, and signal degradation along the way. The reflection properties of the signals incident on these obstacles are not controlled or manipulated/exploited. Nevertheless, conventional wireless communication typically deploys relays and repeaters with their own power supplies, amplifiers, and pre-programmed settings. These active components are placed at static locations to help steer and extend the propagation of the incident signals. In contrast, RIS-aided wireless communication introduces a novel concept by employing passive meta-surfaces with programmable RUs to intelligently manipulate signal reflections in a more flexible and dynamic way. These minimally-powered surfaces can intelligently adjust the phase and amplitude of the reflected signals, allowing for precise control and optimization of the signal paths. This dynamic control enables RIS to steer the incident signals toward the receiver with high accuracy, mitigate interference, and create focused beams, effectively reshaping the wireless environment. This level of control and adaptability is a fundamental departure from the more static and less flexible nature of traditional wireless communication methods.
The control of reflection properties by RIS opens up possibilities for optimizing signal strength and quality at the receiver. An optimal phase shift in RIS contributes to maximizing the number of served devices, improving the SNR, increasing the network sum rate, mitigating signal propagation impairments, and expanding the coverage and capacity in hybrid scenarios [10], [49], [50]. Therefore, optimizing the controllable phase shift towards the receiver becomes crucial for achieving enhanced performance. This has led to significant research interest in beamforming optimization problems for RIS-aided networks. For instance, researchers have explored joint optimization problems involving active beamforming at the BS and passive beamforming at the RIS. Different objectives have been considered, including maximizing the achievable rate or capacity [51], [52], [53], minimizing network power consumption [54], and maximizing EE or SE [55], [56], [57]. Various optimization techniques have been proposed, such as convex optimization [58], alternating optimization (AO) [59], and deep reinforcement learning (DRL) based algorithms [60], [61]. Moreover, specific cases have been examined, such as phase shift optimization for non-ideal RIS with limited phase resolution or uncertain channel information [62], [63].
Furthermore, extensive research has been conducted to evaluate the performance of RIS-aided networks from various perspectives [64], [65], [66], [67], [68], [69]. To assess the performance gains achieved by RIS-aided networks, various common metrics and benchmarks have been used. Some of the key metrics include SNR, bit error rate (BER), outage probability (OP), ergodic capacity (EC), average symbol error probability (ASEP), achievable data rate, coverage area, EE and SE. These metrics help researchers assess and compare the performance of RIS-aided systems under various conditions and scenarios, enabling the optimization and improvement of such systems for different applications. Nevertheless, there has yet to be a benchmark test suite for performance evaluation of RIS-aided networks, as more metrics, such as latency and reliability, are introduced with the emerging demand for ultra-reliable and low-latency communications (URLLC) for 6G wireless systems. Moreover, analytical models and simulation setups have been employed in existing works to evaluate the impact of different design parameters on system performance. These parameters include the number of RUs, BS transmit power, number of users, number of BS antennas, and RIS plane position and location.
Although these studies have yielded promising results, there are still limitations that must be addressed to fully exploit the potential of RIS. Optimizing and evaluating RIS in the context of 6G wireless systems have several challenges and complexities that demand innovative solutions. On one hand, the optimization process involves selecting the optimal phase shifts for each RU to steer reflected signals toward desired user locations. However, in large-scale systems, finding the optimal configuration becomes increasingly difficult, computationally demanding, and time-consuming. On the other hand, the wireless channels in which RIS operates are dynamic and subject to environmental changes, such as user movements and signal blockages. This requires continuous re-optimization of RIS parameters to maintain optimal performance. The dynamic nature of the wireless environment and evolving user needs adds complexity to the optimization process, making it challenging for RIS operation. Nevertheless, RIS-aided systems possess the potential ability to adapt to dynamic environments and evolving user requirements. This adaptability can be achieved through real-time re-configuration, dynamic beamforming, active interference management, and resource-efficient allocation by integrating ML and AI for continuous optimization. Furthermore, evaluating the performance of RIS-aided systems is complex due to numerous factors involved, including dynamic wireless environment, interference, traffic load, and hardware constraints. Consequently, it is crucial to review current state-of-the-art approaches and propose future directions to comprehend these complexities and challenges involved in optimizing and evaluating the performance of RIS-aided wireless systems. This understanding will enable researchers to develop optimal RIS designs that can be practically implemented and evaluated in real-world wireless communication environments.
The objective of this review paper is to provide a comprehensive and in-depth analysis of joint optimization designs and performance evaluation of RIS-aided wireless systems. By thoroughly examining the existing literature and identifying the limitations, challenges, and research gaps, we aim to facilitate future research endeavors in this rapidly evolving field. The insights gained from this review will significantly contribute to the development of optimal RIS deployment strategies and enable the design and optimization of highly efficient RIS-aided wireless communication systems.
It is worth noting that the studies discussed throughout this article primarily focus on optimizing signal reflection and assessing system performance in RIS-aided RF-based wireless systems. Importantly, these studies are not limited to any specific type of RIS materials. RIS technology can indeed utilize various materials and technologies to fabricate the necessary meta-surfaces or meta-materials for manipulating electromagnetic waves effectively. These materials include conductive materials [70], dielectric materials [71], exotic materials like graphene [72], and even plasmonic materials [73]. Additionally, RIS hardware can be composed of meta-structures [74], nanoparticle-based structures [75], near-zero-index materials [76], and multi-functional structures [77], each providing innovative ways to control electromagnetic waves. More in-depth information on RIS hardware can be found in [78]. The selection of materials and technologies depends on specific requirements, operating frequencies, and application goals. While the choice of materials and technologies can impact the performance characteristics of RIS, the optimization strategies and performance assessment approaches discussed here are generally applicable across different material types. However, it’s worth mentioning that the detailed exploration of materials and technologies for fabricating RIS hardware is outside the focus of this paper on joint beamforming optimization design and performance evaluation of RIS-aided networks, rather than the physical hardware of RIS.
D. Related Surveys and Contributions
There have been numerous studies in the literature that have introduced RISs and their applications [11], [40], [78], [79], [80], [81], [82], [83], [84], [85], [86]. However, the focus and scope of these papers differ from our work. Specifically, the authors in [11] extensively investigated the performance of various reinforcement learning (RL) algorithms in RIS-aided networks. In [40], the authors focused on the optimization of RIS phase shifts from the perspectives of signal processing and AI. Specifically, their study provides an overview of different optimization techniques used to optimize the non-convex constrained phase shifts at an RIS. Additionally, it highlights the relationships between these methods and conducts simulations to compare their properties. Moreover, the authors in [78] focused on the hardware aspect of RIS, offering valuable insights for the design of future RIS hardware and providing a solid theoretical foundation. In [79], the authors provided a comprehensive survey on optimization techniques for RIS-aided wireless systems, covering model-based, heuristic, and ML algorithms. A comprehensive analysis of the design and application of RIS in wireless communications is provided in [80]. The authors in [81] presented an extensive exploration of the cutting-edge developments in RIS, including application scenarios, system and channel models, information-theoretic analysis, and essential signal processing techniques. Furthermore, the authors in [82] conducted a comprehensive survey focusing on the research on signal processing techniques that address channel estimation, transmission design, and radio localization issues in RIS-aided systems. Additionally, the authors in [83] provided a comprehensive survey that covers the utilization of RIS in diverse environments, including underwater, underground, industrial, and disaster scenarios. This survey specifically discussed relevant application scenarios, deployment strategies, and various aspects of system design for RIS-assisted communication networks in challenging environments. The authors in [84] directed their attention towards the significance of RIS hardware and system design. They provided an in-depth analysis of various implementation structures of RISs and discussed their utilization through electrical control technologies. In addition, a survey study conducted by the authors in [85] investigated the concept of RISs and the process of channel estimation in RIS-aided networks. Their investigation provided an overview of the model structures of ML and its utilization in various aspects such as channel estimation, spectrum sensing, phase shift, and security. Moreover, the authors in [86] provided a comprehensive investigation of RIS channel estimation principles and approaches, including both traditional methods and advanced AI/ML-based techniques.
Furthermore, several surveys have been conducted to explore the potential benefits of integrating RIS into other emerging technologies [41], [87], [88], [89], [90], [91], [92], [93], [94], [95]. More precisely, the authors in [41] explored the promising potential of using RIS in THz communications, highlighting its application scenarios in wireless mobile communication, UAV, mobile edge computing (MEC), and THz localization. In [87], the authors provided a comprehensive survey that explores a wide range of RIS applications in communication networks involving UAVs. The study also shed light on emerging technologies that have the potential to amplify the benefits of RISs, with a specific focus on both ground and airborne scenarios. The effects of RIS positioning and the roles played by UAVs in enhancing different performance aspects, including SE, EE, reliability, latency, and security, were investigated in a comprehensive survey conducted in [88]. In [89], the authors surveyed the optimization of UAV position and trajectory, as well as precoding and phase shift at the BS and RIS, respectively. Several ML-based techniques for carrying out these optimizations were also presented. In [90], the authors offered a detailed analysis of how UAVs perform when equipped with diverse communication and networking technologies. Moreover, [91] provides a comprehensive explanation of integrating UAVs with IoT technologies. Likewise, the integration of RIS with cutting-edge technologies and applications, such as physical layer security (PLS) and deep learning (DL), has been thoroughly explored in [92] and [93], respectively. The authors in [94] presented a comprehensive review of advanced approaches aimed at addressing the challenges associated with integrating RISs into various emerging multi-user communication technologies, including UAVs, NOMA, mmWave, THz communications, PLS, massive antennas, and SWIPT. Moreover, the authors in [95] provided a comprehensive overview of cutting-edge research on RIS. They focus on RIS operating principles, beamforming design, resource management, applications of ML, and integration with other emerging technologies. Table 1 presents a concise summarization of the existing relevant survey papers on RIS.
Unlike all existing works, this paper offers a critical review of the technical aspects related to joint beamforming optimization and performance analysis in networks aided by RIS. Additionally, our work presents promising research directions for the practical implementation of RIS technology in real-world scenarios. Specifically, our main contributions are as follows:
In contrast to prior research in joint beamforming optimization for RIS-aided networks, which primarily focused on optimizing either the effectiveness of beamforming in terms of rate/capacity or the efficiency of resource consumption in terms of SE/EE, our paper uniquely integrates both. This dual and integrated focus allows for the design of joint beamforming solutions that are not only effective but also resource-efficient.
We provide a comprehensive review of state-of-the-art techniques and approaches for RIS joint beamforming optimization problems across different network configurations and topologies. We synthesize existing literature and identify key advancements, methodologies, and algorithms used to address these problems.
We critically analyze the performance of these techniques, considering their effectiveness, efficiency, convergence properties, and scalability in various scenarios and system configurations.
We highlight current limitations and open research challenges in RIS joint beamforming optimization designs. We offer guidelines for future research directions and inspire the development of novel techniques that can overcome these limitations and further enhance the performance of RIS-aided networks.
We offer a comprehensive and in-depth review of the existing literature on performance evaluation in RIS-aided networks through parametric studies. We synthesize findings from different studies and identify key parameters that significantly influence system performance. This analysis enables an understanding of trade-offs, optimal configurations, and design guidelines for RIS deployment in real-world scenarios.
We address the current research gaps and challenges in analyzing the performance of RIS-aided networks, providing valuable guidance for future research directions and highlighting areas that require further investigation.
E. Paper Structure
This survey paper follows the organization depicted in Figure 3. Section II offers a comprehensive overview of beamforming optimization designs for RIS-assisted systems. Sections III and IV provide a detailed review of joint beamforming design for optimizing system effectiveness and resource efficiency, respectively. Each section is divided into sub-sections based on the application scenarios. Section V explores future research prospects for joint beamforming design, identifying research gaps from Sections III and IV, and emphasizing areas that require further investigation. In Section VI, a critical review of RIS-assisted system performance is presented, with the section divided into two main subsections based on the number of RISs used, each focusing on different application scenarios. Section VII explores open challenges and future research directions for performance analysis and parametric studies, identifying research gaps from the collective studies discussed in Section VI. Finally, in Section VIII, we draw conclusions based on the knowledge and insights derived from this study and outline potential areas for future research extension.
Structure of the paper. Here, “MISO” refers to “multiple-input single-output”, and “MIMO” refers to “multiple-input multiple-output.”
RIS Beamforming Optimization Design
Beamforming optimization is a powerful technique that has the potential to significantly improve the performance of wireless communication systems in RIS-aided networks. Beamforming optimization techniques in RIS-aided networks differ significantly from those used in traditional wireless communication systems. RIS systems can create highly directional beams with precise control over signal direction, which is not possible with traditional omni-directional or sectorized antennas. This makes RIS-aided networks more adaptable to changing user positions and interference patterns compared to traditional systems. Moreover, RIS systems can also actively create nulls in beamforming patterns to reduce interference from specific directions, which is more difficult to achieve with traditional interference cancellation techniques. In addition, RIS-aided systems can optimize EE and SE by focusing signal energy where it is needed. This allows for personalized beamforming for individual users based on their profiles and requirements. Recently, the authors in [47] have successfully implemented an RIS beamforming optimization experiment in a real-world scenario. They demonstrated how the RIS controller runs the beamforming algorithm and optimizes it in real time based on power level feedback. This highlights the software-driven controller’s ability to provide fast real-time optimization, a feature not commonly found in traditional networks. Furthermore, the authors have shown that RIS technology actively enhances signal reception in areas with initially poor signal quality. This is exemplified by the significant power increase achieved through beam adjustment. This result emphasizes the adaptability and precision of RIS-aided beamforming optimization. In contrast, traditional beamforming techniques in non-RIS networks often lack the capability for real-time optimization to this extent. In such systems, beamforming is typically pre-configured and less adaptable to changing conditions. This makes RIS-aided networks more versatile in addressing signal quality challenges.
However, it is essential to note that the optimization algorithms employed in RIS-aided networks can be computationally complex. These complexities can vary based on factors like the scale of the RIS deployment, real-time adaptation needs, channel estimation demands, ML usage, quantization and phase control considerations, and beamforming requirements. Larger RIS systems can lead to more complex optimization problems, requiring longer processing times. Real-time adaptation adds computational intensity, especially when dealing with dynamic environments. Channel state estimation can be computationally demanding, especially in high-mobility scenarios. Implementing ML and AI techniques introduces overhead for model training and real-time decision-making.
Accurate phase control while minimizing quantization errors can also increase complexity. To address these complexities, researchers are developing efficient algorithms to strike a balance between computational demand and optimization performance. Notably, there are several examples of efficient optimization algorithms in RIS-aided networks, as demonstrated by various studies [52], [53], [63], [96], [97], [98]. These algorithms are characterized by their low computational complexity and high performance, providing valuable examples for the ongoing development of efficient solutions in this field.
The effectiveness of RIS beamforming optimization techniques in wireless communication systems can be evaluated using various benchmarks and performance metrics. Some commonly used benchmarks and metrics include SNR, sum rate, achievable rate, BER, SE, EE, ASEP, latency, and coverage. These metrics help researchers to evaluate how well these techniques improve system performance. For instance, when applying a particular beamforming optimization technique, a significant improvement in the achievable rate, or a reduction in BER would indicate the effectiveness of the optimization technique. Furthermore, computational complexity and convergence speed are important performance factors in evaluating the effectiveness of optimization techniques in RIS-aided networks. These metrics have been commonly used by research such as, [51] and [53]. Computational complexity quantifies the computational resources and time needed for an algorithm to execute. Additionally, convergence speed measures the rate at which an optimization algorithm attains an optimal solution. Lower computational complexity and faster convergence speed serve as indicators of the effectiveness of the optimization technique. Nevertheless, there has yet to be a standard benchmark suite for evaluating the effectiveness of these techniques. Effective techniques should strike a balance between achieving optimal performance and being computationally efficient, with a fast convergence rate, to ensure their practicality in real-world scenarios.
However, environmental factors, such as physical obstructions and atmospheric conditions, can impact the effectiveness of RIS beamforming optimization in wireless communication systems. Physical obstructions like buildings and vehicles can obstruct or scatter wireless signals, preventing clear LoS between RIS elements and a user, making it more difficult for RIS elements to optimize signal paths effectively. Atmospheric conditions like rain, fog, or snow can attenuate and scatter signals, making it more difficult for the RIS elements to control the direction of the reflected signal and create a strong enough signal at the user’s location. RIS beamforming optimization algorithms must dynamically adapt to these changing atmospheric conditions to maintain optimal signal paths. Additionally, signal interference from natural sources (e.g., lightning storms) or nearby wireless devices can also degrade the effectiveness of RIS beamforming optimization. Lightning storms can generate electromagnetic interference that can disrupt wireless signals. Nearby wireless devices can also interfere with each other, reducing the signal quality for all devices involved. RIS beamforming optimization algorithms need to be robust such that they can take into account these environmental factors to maintain optimal signal paths and mitigate interference. This can be challenging, especially in dynamic outdoor environments where the wireless environment can change rapidly. Nevertheless, the proposed algorithms for beamforming optimization in RIS-aided networks can have different abilities to handle dynamic environments. Some context-aware algorithms are specifically designed to address these challenges, while others may be more limited in their adaptability. For example, some advanced joint optimization approaches incorporate ML and AI algorithms, such as DRL-based algorithms [99]. These algorithms continuously analyze user patterns and network conditions, making intelligent decisions to optimize RIS configurations based on historical data and real-time feedback. This adaptability ensures the network can meet dynamic environments and rapidly changing user requirements effectively.
Despite the theoretical advancements in optimization techniques for RIS-aided networks, implementing RIS beamforming optimization in real-world scenarios requires several key hardware and infrastructure components. The basis of this technology is the presence of RISs strategically placed in the environment, such as mounting them on a wall or ceiling. These RISs’ elements typically consist of phase shifters, which are essential for controlling the direction of reflected signals. To perform real-time optimization, sufficient computational resources are needed, ranging from general-purpose servers to specialized processing units. These resources handle the complex calculations required for adjusting RIS configurations. Network connectivity is necessary to facilitate communication between RIS devices and controllers, requiring a network infrastructure that supports low-latency communication. Power supplies, even if minimal, are necessary for RIS control and communication. A deployment strategy, ensuring optimal placement and density of RIS elements, is essential for effective beamforming. The deployment strategy should take into account factors such as user locations and wireless environment. Additionally, complex control and management systems, often with security measures, are used to adapt RIS configurations and protect them against unauthorized access or interference. Finally, scalability is essential to accommodate large-scale deployments with numerous RIS elements and users. The infrastructure must support large-scale deployments with multiple RISs and users. These requirements collectively form the hardware and infrastructure needed to exploit the potential of RIS beamforming in practical wireless communication scenarios.
The performance of RIS-aided networks is highly dependent on the optimization of their parameters. In particular, the phase shift at the RIS is a critical parameter that plays a significant role in improving the communication between the source and destination. By adjusting the reflection coefficients of the RIS, it is possible to optimize the phase shift and steer the transmitted signal toward the receiver, thus improving the signal quality and coverage [11]. Besides, optimizing the beamforming at the BS can improve signal strength and coverage, reduce interference, improve spectral efficiency, and minimize power consumption, leading to better network performance and improved user experience [100].
However, the joint optimization design of these two parameters in RIS-aided networks is a complex optimization problem and involves finding an optimal combination of beamforming weights and phase shifts that maximize the system capacity and EE/SE or minimize the total transmission power while ensuring a desired quality of service (QoS) for all users. In this regard, various techniques and algorithms have been proposed in the literature. In the subsequent two sections, we offer a survey of the latest research findings on joint beamforming design along with their significance and limitations. To offer a comprehensive understanding of the current advancements in this field, we have categorized the relevant literature into two distinct groups based on their respective enhancement problems, as presented below.
Joint Beamforming Design for Optimizing System Effectiveness
Numerous articles have explored the joint optimization design of beamforming at the BS and phase shift at the RIS. Both of these factors are crucial in enhancing the user’s rate or system capacity. This section aims to explore studies that primarily focus on optimizing the effectiveness of beamforming designs in terms of user rate or system capacity. In other words, the studies discussed in this section focus primarily on optimal beamforming output in terms of rate/capacity, with less or limited concern about the overall efficiency in the consumption of resources (e.g., power and spectrum). In most cases, power consumption is considered a constraint for optimization rather than an additional optimization objective. We explore problems related to user performance enhancement, including sum rate, weighted sum rate, and achievable rate. Specifically, we review the problem formulations and techniques employed to optimize BS and RIS beamforming for various system scenarios and setups along with their significance, limitations, and potential solutions. For a comprehensive understanding, we categorize the relevant works in this direction into three groups based on their specific application scenarios, as presented below.
A. Single-Cell Multi-User Downlink MISO Scenarios
In this sub-section, we delve into the joint beamforming optimization design for single-cell multi-user downlink MISO scenarios. To provide a comprehensive understanding of this scenario, we have thoughtfully divided this section further into two subsections, based on the considered transmission links. These studies are summarized in Tables 2–3.
1) Single-Cell Multi-User Downlink MISO Scenarios with Direct Link
In [51], the use of RIS in MISO multicast transmission systems is explored in which a multi-antenna BS sends common messages to a group of mobile users with the assistance of RIS. The direct link transmission between BS and each user as well as the RIS-based links are considered to follow Rician fading. It is assumed that the BS and RIS are fully aware of the cascaded CSI of BS-RIS-user links and the CSI of BS-user links. To this end, the authors developed an optimization design to determine the optimal covariance matrix at the BS and phase shifts at the RIS that would maximize the achievable rate. To approach a locally optimal solution, the authors proposed two methods: gradient descent [101] and AO methods [14]. The proposed gradient descent-based algorithm utilizes iterative gradient descent to determine the optimal solution. It begins with an initial guess and then updates the solution by moving in the direction of the steepest descent. The algorithm uses the optimal point from the previous problem in the sequence as the initial value for the current one. During each iteration, the gradient of the objective function is computed for each variable. A step size is then chosen to update each variable in the direction of the steepest descent. This process is repeated until the algorithm converges to the optimal solution. The proposed AO algorithm begins by making an initial guess for the optimal solution. It then proceeds to iteratively update the solution by solving two sub-problems. In each iteration, the algorithm first optimizes the phase shifts using a fixed covariance matrix obtained from the previous iteration. It then optimizes the covariance matrix with a fixed phase shift obtained from the current iteration. The optimization of phase shifts is accomplished using a closed-form expression, while the optimization of the covariance matrix is done using the semi-definite programming (SDP) technique [101]. Through rigorous numerical evaluation and comparison with existing schemes, it has been established that incorporating RIS into the system significantly enhances its achievable rate performance. The proposed algorithms, in particular, have demonstrated remarkable performance, surpassing the beamforming design proposed in [102] and the traditional multicast non-RIS system. In particular, the proposed AO algorithm has outperformed the gradient descent-based algorithm in terms of both performance and convergence speed. Furthermore, the authors analyze the growth order of optimal capacity in some asymptotic cases. Their findings indicate that the optimal capacity increases logarithmically with the number of antennas at the BS and the square of the number of RUs while decreasing inversely with the number of mobile users. Interestingly, when both the numbers of mobile users and antennas approach infinity at a fixed ratio, the optimal capacity remains constant.
The study provided valuable insights into the use of RIS in MISO multicast transmission systems. The proposed algorithms offer effective solutions for optimizing the system’s performance, and the study’s findings can be applied to improve the efficiency and reliability of such systems. Nevertheless, the proposed algorithms have some limitations that should be considered. Firstly, the gradient descent-based algorithm is sensitive to the initial guess for the optimal solution. If the initial guess is far from the true optimal solution, it may converge to a local optimum instead of a global one. Therefore, careful initialization is required to ensure good performance. Secondly, the AO algorithm requires solving two sub-problems in each iteration, which can be computationally expensive. The optimization of the covariance matrix using SDP techniques can be particularly time-consuming for large systems. Thirdly, both algorithms assume ideal continuous phase shifts at the RIS and perfect knowledge of the cascaded CSI between the BS, RIS, and mobile users. In practice, these assumptions may not be applicable. Lastly, both algorithms are designed for locally optimal solutions and may not guarantee global optimality. Therefore, there is a possibility that better solutions exist but were not found by these algorithms. Additionally, the study focused on analyzing the impact of various design parameters on the optimal capacity of the considered system, but it examined them individually rather than concurrently. Therefore, a potential research gap for future investigations lies in exploring the collective impacts of the number of antennas at the BS, the number of RUs, and the number of mobile users. Such research would involve investigating the interrelationships, trade-offs, and their effects on system performance, capacity, and energy efficiency. Furthermore, exploring optimal designs and algorithms that account for the interdependencies between these parameters could offer valuable insights for effectively deploying RIS-based systems in practical scenarios.
The authors of [58] have examined a downlink multi-user MISO system with a RIS-assisted network. In this system, a single BS equipped with multiple antennas serves multiple mobile users with the assistance of RIS. All channels involved, including BS-users, BS-RIS, and RIS-users channels, are assumed to experience Rayleigh fading. For the considered system, the authors proposed a joint optimization approach for active and passive beamforming at the BS and RIS, respectively. The goal was to maximize the WSR of all users while considering practical discrete phase shifts at the RIS and perfect CSI. The formulated problem is decomposed into two sub-problems using the Lagrangian dual transform [103], which allows the active and passive beamforming sub-problems to be solved alternately until convergence is achieved. More specifically, the active beamforming sub-problem is solved using the multi-ratio quadratic transform technique (MRQT) [103], which involves transforming the non-convex problem into a convex one using a set of auxiliary variables and ratios of quadratic forms. The resulting convex optimization problem can be solved with closed-form solutions. In each iteration of the algorithm, the passive beamforming sub-problem is reformulated as a non-convex quadratically constrained quadratic program (QCQP), which is then solved using the alternating direction method of multipliers (ADMM) [104]. ADMM partitions the problem into smaller parts, each of which becomes easier to solve. For comparison purposes, the authors have provided the performance of two baseline schemes. One is non-RIS where the active beamforming is optimized by WMMSE. The other is non-optimized RIS where the phase shifts of the RIS are not optimized but selected randomly. The simulation results demonstrated that the proposed joint beamforming algorithm with a 1-bit phase shift can achieve more than 2 dB performance gain as compared to the non-optimized RIS scheme and more than 4 dB as compared to the non-RIS scheme. Moreover, the study found that a 2-bit quantizer for discrete phase-shifting could achieve a significant capacity gain with only a slight decrease in performance (less than 1 dB) compared to ideal continuous phase shift assumptions. Furthermore, the authors have examined the performance of the proposed algorithm with respect to the number of RUs, showing a positive correlation between them.
The study considered practical limitations on phase-shifting at the RIS and ensured that the proposed optimization strategy is feasible and effective in real-world scenarios. Nevertheless, the proposed approach has some limitations for joint optimization due to the methods employed. One potential limitation of MRQT is that it may not always be feasible to obtain closed-form solutions for the auxiliary variables, particularly for more complex problems [103]. In such cases, numerical methods may be necessary, which can increase the computational complexity of the algorithm. Another limitation of MRQT is that it may not always converge to the global optimum. This is because MRQT entails solving a series of convex optimization problems with varying auxiliary variables, and each solution is dependent on the previous one. If the initial solution is far from the global optimum, MRQT may converge to a local optimum instead. Similarly, ADMM also has its limitations. One such limitation is that it requires solving a sequence of convex optimization problems with linear equality constraints, which can be computationally expensive for large-scale systems [105], [106]. Additionally, the ADMM method can be sensitive to the choice of penalty parameters,
The authors of [63] considered a downlink multi-user MISO with a RIS-assisted wireless network where a multi-antenna BS serves multiple users with the help of RIS. The direct link transmission between the BS and each user, as well as the BS-RIS and RIS-users channels, are considered and modeled using large-scale fading that takes into account the distance between the considered nodes and the shadowing effect. The authors explored the impact of imperfect CSI and channel estimation errors on the performance of the considered system. They utilized a complex normal distribution to model the channel estimation errors for all channels involved. The authors aimed to maximize the average sum rate while adhering to power constraints. To achieve this, they proposed a joint robust design of the active beamforming at the BS and the passive phase shifts at the RIS. To solve this problem, the researchers proposed a new low-complexity algorithm that utilizes the penalty dual decomposition (PDD) technique [108], [109]. The PDD-based algorithm guarantees robustness by simultaneously optimizing the active beamforming at the BS and the passive phase shifts at the RIS to minimize the mean squared error (MSE) of the received signal while accounting for the uncertainty in the CSI. To simplify the algorithm design and to reduce the computational complexity of the algorithm, they have employed the block successive upper bound maximization/minimization (BSUM) method [110]. This has allowed them to reformulate the optimization problem in a more manageable way. The PDD-based algorithm is then used for iterative problem-solving and the attainment of a closed-form solution. The numerical evaluations showed that the proposed algorithm outperforms the existing AO scheme presented in [111] in terms of complexity and sum rate performance. Furthermore, the findings have indicated that the proposed algorithm performs comparably to the ideal CSI scenario. This signifies that the robustness of the proposed design against channel estimation error is good.
The significance of the study lies in its ability to address the challenges posed by channel uncertainty and optimize the system’s performance while minimizing complexity. Nevertheless, it is crucial to recognize that the proposed algorithm is dependent on certain assumptions, such as statistical CSI error models for the cascaded BS-RIS-user and BS-user channels, as well as power constraints. However, these assumptions may not always be relevant in real-world situations, which could potentially affect the efficacy of the proposed algorithm. Therefore, it is imperative to thoroughly evaluate the practical implications of these assumptions before implementing the algorithm in any given context. Moreover, although the proposed PDD-based algorithm is less complex than other methods, it still necessitates iterative procedures to solve the optimization problem in each iteration. Consequently, implementing the algorithm in practice may result in some computational overhead. Furthermore, it would be advantageous to conduct a thorough investigation into the potential impact of phase noise, quantization errors, and other important system parameters such as the number of RUs, RIS placement, and number of BS antennas, on the overall performance of the proposed method. By identifying and addressing these issues, we can ensure that the proposed method is optimized for maximum efficiency and effectiveness, ultimately leading to improved outcomes and results.
The authors in [112] have considered a multi-user downlink RIS-aided MISO femtocell network where the RIS is deployed to assist the communication between one access point (AP) and multiple mobile users. Here, the AP is equipped with multiple antennas while each mobile user has a single antenna. The direct links between the AP and the users are assumed to experience Rayleigh fading while the RIS channels, i.e., channels between AP-RIS and RIS-users are assumed to follow the Rician fading model. To this end, the authors have investigated joint optimization of transmit beamforming at the AP and phase coefficients at the RIS to maximize the weighted sum rate (WSR) of all users subject to the AP transmit power constraint, considering perfect and imperfect CSI. They proposed a low-complexity algorithm for perfect CSI using the fractional programming (FP) technique [103], [113] and block coordinate descent (BCD) method [110]. The FP technique is utilized to break down the initial problem into four separate and distinct blocks. The BCD method is then employed to iteratively update each block of variables while maintaining the other blocks in a fixed state until convergence is attained. This approach allowed for a more efficient and effective solution to the problem at hand. The proposed algorithm is extended to the case of imperfect CSI using the stochastic successive convex approximation (SCA) technique [114], [115]. The SCA algorithm approximates the non-convex objective function using a series of convex functions that depend on the available channel information. By iteratively solving the convex sub-problems and updating the transmit beamforming vector and RIS phase shift vector, the algorithm converges to a stationary point of the original non-convex problem. Besides, the authors provided two baselines for the purpose of comparison. These baselines are, without RIS where the transmit beamforming at the AP is optimized by the weighted minimum mean square error (WMMSE) algorithm [116], [117]. The other baseline is the RIS system with a random phase where the transmit beamforming at the AP is optimized by WMMSE while the phase shifts are not optimized but selected randomly instead. The achieved results illustrated that the proposed joint design schemes significantly outperform benchmarks in terms of performance gain. More precisely, by utilizing RIS with 100 RUs, the joint optimization schemes that have been proposed result in a performance gain of approximately 4 dB when the CSI is perfect, and a gain of approximately 3 dB when the CSI is imperfect for channel certainty level greater than 50%. Moreover, for imperfect CSI, the level of certainty of the channel estimation affects the WSR performance. That is, increasing the uncertainty of channel estimation would decrease the convergence rate of the proposed algorithm. In this context, results have demonstrated that the proposed algorithm exhibits exceptional performance gain, with only minimal degradation observed when the channel certainty estimation is more than 90%. Furthermore, the authors also experimented the proposed algorithm with other design parameters like the number of RUs and the location at which the RIS is placed. In this regard, the result has demonstrated that for fixed transmit power at the AP and fixed location of the RIS, the proposed algorithm yields noticeable performance gains as the number of RUs increases. Additionally, the authors have determined that for a fixed transmit power at the AP and a fixed number of RUs, there is an optimal horizontal distance from the AP to the RIS that results in an optimal performance gain. In this case, the optimal distance is found to be 195 meters. By placing the RIS at this specific location, the wireless signals are maximized, leading to the best achievable performance improvement. This highlights the importance of careful placement of the RIS, as it can significantly enhance the overall system performance.
The study advanced the comprehension of how RIS can enhance wireless network performance and offered a feasible algorithm for practical use. Nevertheless, the proposed algorithms for joint optimization have certain limitations that must be taken into account. One of the primary limitations is the methods employed. For example, FP and BCD techniques may suffer from slow convergence and may not converge to a global optimum in non-convex problems, which can result in suboptimal solutions [103], [118], [119]. On the other hand, the accuracy of SCA is dependent on the degree of non-convexity of the problem [120]. In RIS-aided networks, the non-convexity of the problem can be significant, leading to poor accuracy of the convex approximation and suboptimal solutions. It is important to consider these limitations when selecting an algorithm for joint optimization in RIS-aided networks. By doing so, we can ensure that we are using the most appropriate algorithm for the specific problem at hand, and ultimately achieve optimal results. Moreover, investigating the impact of channel estimation errors on the performance of RIS-aided systems and developing more robust algorithms that can handle imperfect CSI cases with higher uncertainty levels could help improve the reliability and performance of the system in real-world scenarios. Furthermore, investigating the practical implementation issues associated with deploying RIS in real-world networks, such as hardware constraints, power consumption, and scalability could help identify and address the challenges of implementing RIS technology in practical settings. In addition, although the authors experimented with the number of RUs and the location of the RIS in their study, they did not explicitly investigate the combined impact of these factors on system performance. Understanding how the number of RUs and the placement of the RIS interact and influence the overall performance of the system could be an interesting area for further research. Table 2 provides a summary of the studies discussed in this section to facilitate better comprehension.
2) Single-Cell Multi-User Downlink MISO Scenarios With Blocked Direct Link
Researchers of [62] have considered a multi-user downlink RIS-assisted MISO system where a single BS equipped with multiple antennas serves multiple mobile users with the help of RIS. The direct link transmission between the BS and each user is assumed to be blocked while the RIS channels (i.e., BS-RIS-users channels) are assumed to follow the Rician fading model. To this end, the authors have proposed a hybrid beamforming (HBF) design that utilizes perfect CSI assumption and practical discrete phase shifts at the RIS to maximize the users’ sum rate. The HBF design is comprised of two parts: a continuous digital beamforming part executed at the BS and a discrete analog beamforming part performed at the RIS. The authors developed an iterative algorithm called the sum rate maximization (SRM) algorithm to solve the digital/analog beamforming sub-problems alternatively. Specifically, the SRM algorithm works in an iterative manner where the digital beamforming part at the BS is solved using zero-forcing (ZF) with power allocation [121], while the analog part at the RIS is solved by the outer approximation method [122]. To provide a comprehensive comparison, the authors also evaluated several other algorithms, such as the simulated annealing method [123] for approaching the global optimal solution, the random phase shift algorithm for solving analog beamforming sub-problem, and relaxation-based optimization [124]. Results showed that the proposed algorithm performs comparably to the simulated annealing method, and significantly surpasses both the random algorithm and the relaxation-based algorithm in terms of performance gain. Additionally, the authors have provided theoretical analysis and numerical results to explore the impact of the number of RUs and the number of quantization bits on the achievable rate performance. In this context, results showed that by utilizing discrete phase shifts, the sum rate of the system can increase rapidly with only a small number of quantization bits. If the number of quantization bits is sufficiently large, the sum rate gradually gets close to the one achieved in the continuous scenario. Moreover, as the number of RUs increases, the sum rate also increases and eventually stabilizes when the number of RUs is 40. Finally, results illustrated that the RIS-based HBF scheme required only half the number of transmit antennas at the BS compared to traditional HBF schemes. This implies that utilizing the RIS-aided HBF technique conveys a substantial decrease in the cost required for dedicated hardware. These observations suggest that a reasonable number of RUs, several BS antennas, and a small number of quantization bits are sufficient to achieve a satisfying sum rate at a low cost when designing RIS-based systems. Therefore, it is important to consider these factors when designing RIS-based systems to ensure optimal performance and cost-effectiveness.
The authors proposed a practical and efficient RIS-based HBF scheme for multi-user communications, which can achieve high achievable rates with a limited number of discrete phase shifts. Nevertheless, it is advantageous to explore the impact of hardware impairments on the functionality of the proposed design. By doing so, we can gain a better understanding of how these impairments may affect the overall performance of the system. This knowledge can help identify potential issues and develop strategies to mitigate them, ultimately leading to a more robust and reliable design. Furthermore, the authors acknowledged that the correlation between the RUs at the RIS can negatively impact data rate performance. Therefore, it is imperative to investigate the impact of this correlation on data rate performance and develop techniques to mitigate its effects. Finally, there are some limitations in the proposed algorithm due to the methods employed. For instance, while ZF with power allocation can achieve a perfect balance between power allocation and user rates, resulting in an efficient and effective system, it assumes perfect CSI for all users. This may not be available in practical applications, which can limit the effectiveness of the proposed algorithm. On the other hand, the outer approximation method can handle complex constraints, such as discrete phase shifts for RIS, while still achieving optimal performance. However, the downside of this technique is that it can be computationally expensive, particularly for large-scale systems [62]. Thus, careful consideration should be given to these limitations and further research is needed to explore the potential of this scheme in practical applications.
In [96], the authors have considered a RIS-aided multi-user downlink MISO communications system where a BS with multiple antennas serves multiple users with a single antenna via an RIS. The direct link transmission between the BS and each user is assumed to be unavailable due to blockage. It is assumed that the RIS-based channels are Rician faded channels. For the considered system, the authors develop a low-complexity joint optimization design of active beamforming at BS and passive phase shifts at RIS to maximize the sum rate of all users under both perfect and imperfect CSI scenarios. For perfect CSI, the formulated non-convex optimization problem has been tackled by subdividing the original problem into three different and distinct sub-problems using FP transformations. By introducing auxiliary variables in each sub-problem, a closed-form expression of the beamforming vector and reflecting matrix is derived resulting in a low-complexity joint beamforming and reflecting alternating algorithm. For imperfect CSI, the authors adapt the proposed algorithm and develop a corresponding design with channel estimation error. The performance of the proposed algorithm has been evaluated numerically in terms of achievable sum rate and complexity. To compare, the authors have also evaluated the performance of the existing schemes in the literature such as ADMM and MMSE/ZF+ semi-definite relaxation (SDR) [105] based schemes. Through extensive simulation, the results validated the effectiveness of the proposed algorithm in terms of both complexity and achievable sum rate over the aforementioned benchmark schemes even in the presence of channel uncertainty. Moreover, the study investigated the rate performance of the proposed algorithm with various system parameters, including the number of RUs, the number of antennas at the BS, the number of users, and the horizontal distance between the BS and RIS. The findings revealed that the rate performance of the proposed algorithm increases as the number of RUs or BS antennas increases, while it decreases with an increase in the number of users. As for the impact of the RIS placement, the results indicated that the optimal distance from the BS for which the RIS should be placed to achieve the best rate performance is 190 meters. It is worth noting that the study investigated the impact of these design parameters individually, meaning that one parameter was changed at a time while keeping others fixed. However, it would be valuable to explore the joint impact of these parameters on the performance of the proposed algorithm to gain more insights into system performance and optimization.
The proposed algorithm has the potential to significantly reduce computational complexity, making it a more practical option for implementation in RIS-aided systems. Nevertheless, the proposed algorithm has some limitations that are associated with the methods employed. For instance, achieving convergence to the global optimum can be challenging in FP-based algorithms. The presence of non-convex and non-linear fractional objective functions may lead to multiple local optima, making it difficult to guarantee global optimality. Additionally, auxiliary variable methods can require solving a series of optimization sub-problems. This can be computationally expensive, particularly for large-scale problems, as each sub-problem may involve non-convex or complex constraints. It is crucial to note that these limitations do not necessarily render the proposed algorithm ineffective. Rather, they highlight the need for careful consideration and evaluation of the algorithm’s performance in specific contexts. By acknowledging these limitations, researchers can work to develop strategies for mitigating their impact and improving the algorithm’s overall effectiveness. Furthermore, investigating the performance of the proposed method in more complex scenarios, such as multi-cell and multi-user scenarios will provide a more comprehensive understanding of the method’s capabilities and limitations in diverse settings.
The authors in [97] have considered a multi-user downlink MISO with a RIS-aided network where a BS equipped with multiple antennas serves multiple users via an RIS. The direct link transmission between the BS and each user is assumed to be blocked while the BS-RIS and RIS-users channels are assumed to follow the Rician fading model that takes into account the LoS links, the NLoS links, and the distance-dependent path loss. With perfect CSI assumption, the authors have proposed a joint beamforming design of the active beamforming at the BS and the passive beamforming (phase shift) at the RIS so that the sum rate of all users can be maximized. The study argued that the widely used iterative optimization algorithms that rely on sub-optimal solutions are not suitable for practical scenarios due to their high computational complexity. To address this issue and reduce the computational complexity while maintaining comparable sum rate performance, the authors have proposed a DL-based approach for joint active and passive beamforming design. The proposed approach involved training a two-stage neural network offline in an unsupervised manner before deploying it online for real-time prediction. The first stage of the network predicts the phase shifts to form the optimal effective channel matrix, while the second stage predicts transmit beamforming based on the predicted channel matrix. The performance of the proposed DL-based approach is evaluated numerically in terms of complexity and sum rate. To provide a comprehensive comparison, the authors compared the proposed approach with baseline schemes used in the literature such as the iterative optimization algorithm proposed in [112] and random phase shifts plus WMMSE beamforming. Results have demonstrated that the proposed approach can achieve comparable sum rate performance with significantly less computational complexity. Specifically, the proposed algorithm has been proven to be more efficient than the iterative algorithm in certain scenarios. It performs exceptionally well when the number of RUs is 64 or greater and the number of iterations in the iterative algorithm is 100. Even when the number of RUs varies, it still achieves over 93% of the iterative algorithm performance with 500 iterations. This highlights the superiority of the proposed algorithm and its potential to be a valuable tool in various applications. However, the study only considered a specific scenario of RIS-aided multi-user MISO downlink systems, which raises questions about its applicability in other contexts, such as multi-user MIMO systems. Additionally, the proposed approach required offline training of a two-stage neural network in an unsupervised manner, which may require significant computational resources. While the proposed approach may still be too computationally expensive for some practical scenarios, it shows promise for enhancing wireless network performance in RIS-aided communication systems. Further research is needed to address these limitations and evaluate their effectiveness in real-world scenarios.
The work in [99] has considered multi-user downlink MISO with a RIS-aided network where one BS equipped with multiple antennas serves multiple users via RIS. Each user is equipped with a single antenna and the communication is only performed through RIS (i.e., the direct link transmission between the BS and each user is assumed to be blocked). The RIS-based channels (i.e., the BS-RIS and RIS-users channels) are assumed to follow Rayleigh distribution, and perfect CSI at the BS and RIS is assumed. To this end, the authors have proposed a joint optimization design of the transmit beamforming at the BS and the phase shifts at the RIS that maximizes the user’s sum rate. To solve the formulated non-convex problem, the authors have developed a DRL [60], [61] based algorithm that can learn from the environment and gradually improve its behavior. The proposed DRL algorithm uses a powerful neural network to determine the best way to transmit data. It takes into account the current state of the environment and uses this information to make decisions about the transmit beamforming matrix and phase shifts. The performance of the proposed DRL-based algorithm in terms of average sum rate has been numerically evaluated and compared with two state-of-the-art benchmark schemes. The selected benchmark algorithms are the WMMSE algorithm, and the iterative algorithm based on FP with ZF beamforming [125]. The results showed that the proposed DRL-based algorithm significantly outperforms the aforementioned benchmark algorithms. Furthermore, the authors have observed that increasing the transmission power at the BS and the number of RUs significantly improves the performance of the proposed algorithm. This highlights the importance of optimizing these parameters in such systems.
The results of this study indicated that the proposed algorithm has the potential to significantly improve the performance of wireless communication systems. One of the most significant advantages of the proposed DRL is its ability to learn optimal policies for RIS-aided systems. This is particularly important in dynamic environments where the system’s performance can change rapidly. DRL can adapt to these changes and optimize the system’s performance in real-time, leading to enhanced network efficiency and user experience. Moreover, the proposed algorithm does not require sophisticated mathematical formulations or numerical optimization techniques, which can be computationally expensive and time-consuming. Instead, it uses a neural network to approximate the optimal transmit beamforming matrix and phase shifts, which can be trained using data obtained from interacting with the environment. This makes it a valuable tool for future wireless communication systems. Nevertheless, it is worth noting that the effectiveness of the proposed algorithm may depend on the quality and quantity of data used for training the neural network. Moreover, the utilization of neural networks for approximation in DRL algorithms does not provide any interpretability, and as a consequence, the resulting algorithm may converge to suboptimal solutions or get stuck in local optima [99]. Although DRL algorithms can be effective in finding promising solutions, there is no guarantee of achieving the globally optimal solution. Additionally, the study assumed perfect CSI and developed the optimization frameworks accordingly. However, it’s important to note that obtaining perfect CSI is often challenging in practice due to various factors such as channel conditions, noise, and interference. Therefore, further research is needed to explore the effectiveness of the proposed design in imperfect CSI scenarios and to optimize the quality and quantity of data used for training the neural network. For a better understanding of this section, we have summarized the studies discussed in Table 3.
B. Single-Cell Single-User Downlink MIMO Scenarios
A multi-stream downlink RIS-aided MIMO system is analyzed in [52]. The considered system involved multiple antennas transmitter that communicates with multiple antennas stationary receiver with the assistance of RIS. The direct link transmission is assumed to be blocked and the RIS-based links are assumed to follow the Rician fading model. By considering perfect CSI, the authors proposed a joint optimization design for the covariance matrix of the transmitted signal and phase shift at the RIS to maximize the achievable rate under transmit power constraint. To achieve this, the authors have utilized the iterative projected gradient (PGM) method [126] where all optimization variables are updated simultaneously in each iteration rather than alternately. The proposed algorithm iteratively updates the phase shift vector and covariance matrix values based on their gradients and step size. This process involves projecting them onto feasible sets to ensure compliance with the transmitted power constraint. The algorithm continues to iterate until convergence is achieved, ultimately resulting in the optimal values for these variables. The proposed algorithm’s performance has been evaluated numerically in terms of achievable rate and complexity and compared with the widely used AO algorithm. It is worth mentioning that the complexity of the proposed algorithm has been measured based on the number of iterations that the algorithm needs to achieve optimal rate performance. The results showed that the proposed algorithm achieved the same achievable rate as the AO algorithm but with significantly less complexity. It achieves a remarkable 98% reduction in the number of iterations required when compared to the AO method. The study also examined the resilience of the proposed algorithm to real-world system imperfections, including imperfect CSI and discrete phase shifts at the RIS. In this context, results showed that even with a very low resolution of discrete phase shifts, the optimal achievable rate offered by the proposed algorithm is only slightly reduced compared to continuous phase shifts. In particular, the use of 1-bit discrete phase shifts results in a reduction of approximately 1.1 bit/s/Hz, while 2-bit discrete phase shifts lead to a reduction of approximately 0.2 bit/s/Hz. Moreover, the study also found that in the case of imperfect CSI, the optimal achievable rate decreases by approximately 1 bit/s/Hz compared to perfect CSI. This reduction is considered acceptable. These results demonstrated that the proposed algorithm exhibited exceptional performance even in the presence of these challenges, providing valuable insights into the feasibility of the suggested design. Additionally, the authors examined the achievable rate with respect to the number of RUs at the RIS. In this regard, they show that as the number of RUs increases, the achievable rate also increases, with a larger increase observed when the number of RUs is doubled. Although the authors provide valuable insights into the impact of the number of RUs on the performance of the proposed design, it is important to note that other critical system parameters, such as user mobility, RIS placement, and the number of antennas at the BS and receiver, are not considered. As a result, further investigation is required to comprehensively understand the implications of these factors.
The proposed algorithm offers a promising solution to the optimization of RIS-assisted wireless communication systems. Its exceptional performance in the face of real-world imperfections and reduced complexity make it a valuable addition to the existing literature. However, there are some limitations associated with the proposed algorithm that must be considered. Firstly, the PGM may converge to a local optimum instead of a global optimum due to the objective function’s multiple local optima. Secondly, the PGM requires tuning of hyper-parameters such as step size and projection parameters, which can be challenging and require careful consideration. Lastly, the proposed algorithm assumes that all antennas have equal power constraints, which may not be the case in practice. The optimization of power allocation among different antennas with unequal constraints is a challenging problem that requires further investigation. Further research is needed to address these limitations and optimize the algorithm’s performance in various scenarios.
The authors in [53] have considered a downlink RIS-assisted MIMO system consisting of a BS equipped with multiple antennas, a single stationary user equipped with multiple antennas, and RIS. The direct link transmission between the BS and the user as well as the RIS-based links are all assumed to follow the Rician fading model. For the considered system, the authors introduced an integrated approach to optimizing the transmit covariance matrix at the BS and the diagonal phase shifts at the RIS to maximize the achievable ergodic rate subject to the transmit power constraint at the BS. This approach is based on statistical CSI at the BS and employs a large-scale system approximation and AO algorithm. In particular, an analytical expression for the achievable ergodic rate in the large-system regime has been derived using the replica method [127]. The obtained large system approximation is then used to determine the asymptotic optimal transmit covariance matrix at the BS and the optimal diagonal phase shifting at the RIS using an AO algorithm. The use of the large-system approximation can potentially reduce the computational complexity of optimizing these matrices in large-scale systems. This is because the approximation is valid for large values of the number of antennas, users, and RUs, which are common in practical wireless communication systems.
The proposed AO algorithm is designed to optimize these matrices separately. This approach starts with an initial guess for one of the matrices and then alternates between updating one while keeping the other fixed until convergence is achieved. The transmit covariance matrix is designed using the water-filling technique [62], [128]. This technique involves allocating power efficiently across transmit antennas based on the statistical CSI. It assigns higher power to antennas with better channel conditions and reduces power for antennas with weaker channels, aiming to maximize the achievable rate while adhering to power constraints. The diagonal phase shifts at the RIS, on the other hand, are optimized using the projected gradient ascent method [129]. This technique iteratively adjusts the phase shifts to maximize the achievable rate. The method begins with an initial guess for the phase shifts and updates them iteratively based on the gradient of the objective function. The gradient represents the direction of the steepest ascent towards the maximum value. To ensure that the phase shifts remain within a feasible range, a projection step is performed at each iteration. This projection maps the updated phase shifts back into the feasible region if they exceed the boundaries. The process continues until convergence is achieved. Numerical results validate the effectiveness of the proposed algorithm and show a performance improvement of the considered RIS-aided MIMO system as compared to traditional MIMO without RIS and conventional MIMO with amplify and forward (AF) relay technology [130]. Additionally, the authors have assessed the achievable ergodic rate provided by the proposed algorithm with respect to the number of iterations. It has been demonstrated that the proposed algorithm achieves optimal rate performance in 25 iterations, which confirms the convergence of the proposed algorithm. Furthermore, the authors evaluated the proposed algorithm’s achievable ergodic rate performance in relation to the number of RUs. The results demonstrate that the achievable ergodic rate increases as the number of RUs increases. Moreover, the study highlights the impact of RIS placement on the achievable rate, showing that placing the RIS closer to the BS or user results in higher rates, while placing it in the middle leads to lower rates. These findings offer valuable insights that should be carefully considered when deploying RIS in real-world scenarios. It is important to take into account the implications of these findings to ensure successful implementation and optimal outcomes.
The implementation of large-scale approximation and AO algorithms can provide significant benefits to the design and optimization of the system under consideration, particularly in reducing the computational complexity of optimizing large-scale systems. However, it is important to be aware of certain challenges and limitations, such as the high computational cost of solving two sub-problems, the sensitivity to the initial guess or starting point for the phase shifts, as well as the sensitivity to noise and interference, and the possibility of becoming trapped in local optima. These challenges must be carefully addressed to ensure the optimization process is both effective and efficient. Moreover, the authors have emphasized the impact of various system parameters, such as the number of RUs, and placement of RIS, on the rate performance of the proposed algorithm. However, it is crucial to consider other essential design parameters in future research, including user mobility, the number of antennas at the BS, and the number of antennas at the end-user. Without this knowledge, the research community may not have a complete understanding of the potential impact of these design parameters on the achievable performance gain. Therefore, it is imperative to conduct further studies to explore the effects of these parameters and their potential contributions to the algorithm’s overall performance. Table 4 provides a summary of the studies discussed in this section.
C. Other Scenarios
In the following, we review the recent state-of-the-art in joint beamforming design for optimizing system effectiveness in other considered scenarios. To ensure a clear and comprehensive understanding of this section, we have summarized these studies in Table 5.
The authors of [98] investigate a downlink multi-user MIMO with a RIS-aided system that takes into account practical hardware constraints such as limited phase shift resolution at the RIS. The considered system consists of one BS with multiple antennas that serve with the assistance of RIS multiple users each equipped with a single antenna. Distance-dependent path loss model is used to represent the large-scale fading for direct and RIS-based channels while Rician fading is considered to model the small-scale fading of the direct links. The study assumes perfect CSI which allows the BS to utilize the classical ZF as the active beamforming scheme for signal transmissions. To this end, the authors proposed a joint optimization design of the active beamforming at the BS and passive beamforming at the RIS that maximizes the achievable sum rate. To tackle the formulated maximization problem, the authors proposed a low-complexity biologically inspired particle swarm optimization (PSO) algorithm that operates in a stochastic iterative manner. The proposed PSO algorithm operates in three distinct steps. Firstly, several particle swarms, which represent possible passive beamformers, are generated with random positions that are normalized to ensure unit power. Secondly, the velocity of each particle is randomly initialized, and the active beamforming matrix can be obtained from the equivalent channel matrix. Finally, the objective function, which is the achievable sum rate, is computed for each particle. The particle that maximizes the objective function is considered the best particle in one iteration, and its velocity and position are updated for the next iteration. In subsequent iterations, each particle keeps track of its own best position, and new objective function values are updated accordingly. This process is repeated until the maximum number of iterations is satisfied. The best particle with the highest objective function value is regarded as the optimal solution. Numerical evaluation reveals that the proposed PSO-based algorithm can attain 93% of the SE achieved by the AO scheme proposed in [142], at lower computational complexity. This makes it a promising option for real-time implementation in practical systems.
While the proposed scheme is shown to be computationally efficient and achieves high performance, it does not necessarily guarantee fairness or QoS for all users. This is an important consideration in practical systems where multiple users with different requirements may be competing for network resources. Additionally, PSO is an iterative algorithm that can be computationally expensive and time-consuming for large-scale systems. It is also susceptible to being trapped in local optima and may not guarantee to find the global optimum. Moreover, the performance of the proposed PSO algorithm may be affected by the choice of parameters, such as the number of particles and the maximum number of iterations. Furthermore, the study assumes that CSI is perfectly known at both the BS and RIS. This assumption may not hold in practice due to channel estimation errors or feedback delays. Therefore, investigating the impact of imperfect CSI on the performance of the proposed algorithm is an important consideration for practical implementation. In addition, extending the proposed scheme to consider other performance metrics besides sum-rate maximization, such as fairness or energy efficiency, can be suggested to enable a more comprehensive evaluation of RIS-aided MIMO systems in different scenarios and applications. Finally, the study has a notable limitation in that it fails to account for the impact of system design parameters on the performance of the proposed algorithm. These parameters include the number of RUs and placement of the RIS, the number of antennas at the BS, and the number of users. To gain a more comprehensive understanding of the algorithm’s performance, further research is required in this area.
In [131], the authors studied a multi-cell multi-user downlink MIMO with a RIS-assisted wireless network. The study considers a two-cell scenario where each cell is equipped with a single BS that utilizes multiple transmit antennas to serve two users. Each user has multiple receive antennas, receiving attenuated signals from its serving BS and experiencing co-channel interference from neighboring BS. As depicted in Figure 4, one RIS is deployed at the cell edge of the two neighboring cells to assist downlink transmission by adjusting the phase shift of the RUs, which helps to alleviate interference. Both the direct link transmission from the BSs to users as well as the transmission links from BSs to RIS and from the RIS to users are considered where the distance-dependent path loss is used to model the large-scale fading for all involved channels. However, Rayleigh and Rician fading models are used to model the small-scale fading of the direct and RIS-based channels, respectively. By assuming perfect CSI, the authors proposed a joint optimization design of transmit precoding matrices of all BSs and the phase shifts at the RIS that maximizes the WSR of all users subject to each BS’s power constraint and the unit modulus constraint of the phase shifters. The authors have proposed a BCD algorithm to solve the problem in an alternating manner. The optimal precoding matrices were obtained in closed form using the Lagrangian multiplier method [132], while two efficient algorithms to address the phase shift optimization problem, called majorization minimization (MM) algorithm [133] and complex circle manifold (CCM) algorithm [134], were presented. In particular, the BCD method was utilized to update each block of variables iteratively, while keeping the other blocks in a fixed state until convergence is achieved. During each iteration of the BCD algorithm, either the MM or CCM algorithm is employed to determine the phase shifts of the RIS. This resulted in two different optimization algorithms: BCD-MM and BCD-CCM. Besides, for comparison purposes, the authors provided two benchmark schemes to select the phase shifts at the RIS while the optimal transmit precoding matrices at the BSs are obtained using the BCD algorithm in both cases. These benchmarks are the random approach scheme where the phase of each RU is randomly generated in a uniformly distributed manner and the No-RIS scheme where the RIS channels matrices are set to be zero. The results obtained indicate that the proposed BCD-MM and BCD-CCM algorithms have achieved nearly identical WSR and have shown to be highly effective and efficient in improving the performance of multi-cell systems as compared to the benchmarks. In addition, results illustrate that incorporating RIS at the cell boundary can minimize inter-cell interference and improve the quality of communication for users who are located at the edge of the cell. This has important implications for improving wireless systems and meeting the increasing demand for high-speed and reliable wireless connectivity. Moreover, the study assessed how the number of RUs affects the performance of the proposed algorithms. The results demonstrated that an increase in the number of RUs leads to a substantial improvement in performance. Specifically, when there are 10 RUs, the performance gain compared to the No-RIS scheme is 2 bit/s/Hz. However, this performance gain significantly increases to 13 bit/s/Hz when there are 80 RUs. Furthermore, the authors extended the algorithm to a four-cell scenario, in which they examined two different approaches for deploying RIS. The first approach is a centralized deployment, where a single RIS with 50 RUs is positioned at the cell edge between the first two cells. The second approach is a distributed deployment, where two RISs are utilized, each equipped with 25 RUs. In the latter approach, one RIS is placed at the cell edge between the first two cells, while the other is placed at the cell edge between the remaining two cells. These cases are illustrated in Figure 5 for clarity. The results demonstrated the superiority of the distributed deployment strategy over the centralized deployment strategy. This could have important implications for improving wireless communication systems beyond multi-cell MIMO.
![FIGURE 4. - Illustration of the two-cell scenario considered in [131].](/mediastore/IEEE/content/media/6287639/10005208/10356072/mahmu4-3342320-large.gif)
![FIGURE 5. - Illustration of the different approaches for deploying RIS in the four-cell scenario considered in [131].](/mediastore/IEEE/content/media/6287639/10005208/10356072/mahmu5-3342320-large.gif)
Illustration of the different approaches for deploying RIS in the four-cell scenario considered in [131].
The study advances the comprehension of how RIS can enhance wireless network performance and offers efficient and effective algorithms for solving complex problems. However, it is important to consider the limitations of the proposed algorithms due to the methods employed. For instance, the MM and CCM algorithms used in each iteration of the BCD algorithm for finding the phase shifts of the RIS require an increasing number of iterations for convergence as the number of phase shifts increases. This means that the complexity of the proposed BCD-MM and BCD-CCM algorithms may increase in practical scenarios. Consequently, the scalability of the proposed algorithms to large-scale RIS systems, with a large number of RUs and antennas, may pose challenges. Moreover, both algorithms can converge to a local optimum but may not guarantee global convergence to the optimal solution due to the non-convex nature of the optimization problem. It’s crucial to keep these limitations in mind when implementing these algorithms in real-world situations. Furthermore, assuming perfect CSI and ideal hardware may not hold in all scenarios, leading to an overestimation of the technique’s performance. Future research could investigate how to overcome these limitations and improve its practicality and performance. Although the authors have investigated the impact of the number of RUs on the performance of the proposed algorithms, other design parameters, such as the number of antennas at the BS and for the users, and the number of users, have yet to be investigated. Therefore, further research is necessary to explore these factors and their joint potential impact on the proposed algorithms.
The authors in [135] explore an uplink mMIMO with a RIS-assisted communication system in the presence of a direct link. The considered system consists of multiple mobile users each of which is equipped with a single antenna that communicates with multiple antennas BS with the assistance of RIS. The direct links between users and BS are assumed to follow the Rayleigh fading model while the users-RIS and RIS-BS channels are assumed to be Rician faded channels. By employing the maximum ratio combining (MRC) technique [136] and assuming perfect estimation of the CSI using the same estimation method as in the traditional mMIMO system [137], the authors derived a closed-form expression for the uplink ergodic data rate of the considered system. The derived expression is then used to find the optimal phase shift at the RIS that maximizes the user sum rate using a genetic algorithm (GA) [138], [139]. The proposed GA-based method for designing the phase shifts of RIS works by starting with a population of 200 individuals, each with a randomly generated chromosome. The fitness of each individual is calculated using a sum data rate maximization problem. The top 10 individuals with higher fitness are kept as elites and 40 individuals with lower fitness are replaced with new individuals generated using uniform mutation [139]. The remaining individuals are selected for crossover to generate new offspring, which are added to the population. This process continues until a maximum number of generations is met. Finally, the individual with the highest fitness value among all generations is selected as the final solution. The performance of the proposed GA-based method has been evaluated numerically and compared with conventional mMIMO without RIS and mMIMO with RIS employing random phase shifts. Results showed that, at low transmission SNR, the optimized RIS mMIMO system outperforms the aforementioned conventional systems. Further, at high SNR regions, the optimized RIS-based system still performs better than the others. However, at extremely high transmission SNR regions, the conventional mMIMO system outperforms the optimized system. Thus, it can be concluded that the proposed RIS-based system can play a significant role in the low to high SNR regime though not at extremely high SNR. Moreover, the authors have studied the performance of the considered systems with respect to the number of BS antennas and the number of RUs. It has been shown that as the number of BS antennas or the number of RUs increases, the RIS with optimal phase shifts brings a significant performance improvement over other systems. However, in scenarios where the direct links are strong or when the number of BS antennas is large, mMIMO systems without RIS have a better performance than RIS-aided systems with random phase shifts. These results suggest that RIS can be an effective technology for enhancing the performance of mMIMO systems, but their implementation needs to be optimized for different scenarios and system configurations.
Considering statistical CSI allows the authors to design the phase shifts of the RIS based on a more realistic and practical model of the wireless channel. In addition, using MRC provides a low-complexity beamforming technique that combines the signals received from multiple antennas. Therefore, using statistical CSI and MRC, the complexity and overhead of the system can be reduced while still achieving good performance. Thus, the proposed approach has important implications for the practical implementation of RIS-aided mMIMO systems, as it provides a more efficient and effective way to design and optimize the phase shifts of the RIS. However, the proposed GA-based method may have some computational complexity, particularly when the search space is large, which may increase the computational cost and time required for the optimization process. Additionally, the effectiveness of the GA algorithm is influenced by the appropriate selection of its parameters, such as the mutation probability. Hence, careful consideration and tuning of the parameters are crucial for achieving high-quality solutions. Furthermore, the authors assume that the RIS has perfect phase shift control, which may not be achievable in practice due to errors or delays in adjusting the phase of the reflected signal. Additionally, the authors assume that CSI is only available at the BS and is estimated using low-overhead statistical techniques, which may not be sufficient to achieve optimal system performance in practical scenarios. Therefore, it is important to keep these limitations in mind when interpreting the achieved results and applying them to practical scenarios. Finally, it is worth mentioning that the study lacks consideration for other design parameters, such as RIS placement, and the number of users. Further research is necessary to address the impact of these parameters on the performance of the proposed algorithm and explore the interdependencies between them.
The authors in [140] investigated a downlink multi-user single-input single-output (SISO) communication system where multiple pairs of mobile users, each having a single antenna, communicate through an RIS. The direct link transmissions between users are assumed to be blocked while the RIS-based links (i.e., the transmit user-RIS and RIS-received user channels) are assumed to follow the Rician fading model. The study assumes statistical CSI availability at the transmitting user and derived a closed-form approximation for the ergodic achievable rate at the received user. Based on the analytical framework and by considering both continuous and discrete phase shift setups at the RIS, the optimal phase shift that maximizes the sum achievable rate is obtained using the GA method. The proposed GA algorithm’s effectiveness and the analytical framework’s correctness have been numerically verified and compared to a scheme based on randomly chosen phase shifts and a solution obtained through the exhaustive search scheme [141]. The results demonstrate that the GA-based design outperforms the RIS-based system employing random phase shifts and performs nearly the globally optimal solution obtained through the exhaustive search method. This observation is particularly intriguing and highlights the potential of the proposed GA method in achieving optimal results. Moreover, their finding demonstrated that using GA and a three-bit quantization can significantly improve the achievable rate for the discrete phase shift setup and can achieve a significant portion of the total achievable rate for the continuous phase shift setup. This provides useful engineering insights for designing RIS-aided systems, as it suggests that a low-resolution phase shifter may be sufficient for practical implementations. Furthermore, using discrete phase shifts also enables the use of digital signal processing techniques, which are more efficient and easier to implement than analog techniques. This can lead to significant cost savings and improved performance in practical systems. However, it’s important to note that using discrete phase shifts also introduces quantization errors, which can degrade system performance. The authors addressed this issue by proposing a GA method that can optimize the phase shifts with a small number of bits while minimizing quantization errors. Nevertheless, the proposed GA method can be computationally expensive and time-consuming for large-scale optimization problems, which could limit its usefulness in practical applications. In addition, the assumptions made in the study may not hold in all practical scenarios, which could limit the real-world applicability of the proposed method. For instance, obtaining accurate statistical CSI may be challenging due to various factors such as hardware limitations and environmental conditions. Moreover, assuming independent channels and fixed transmit power simplifies the problem and reduces its complexity. However, this may not always be realistic and could limit the generalizability of the proposed method. As a consequence, it is essential to be mindful of these constraints when assessing the findings and putting them into practice.
Joint Beamforming Design for Optimizing Resource Efficiency
As described in the previous subsection, the use of RIS in wireless networks has gained significant attention due to its ability to enhance signal strength and system capacity. However, to fully exploit the benefits of RIS, it is crucial to delve into joint beamforming designs that can help minimize power consumption or maximize the EE/SE of the system while still maintaining high signal quality and efficient transmission. This can help reduce the network’s operating costs and increase overall network performance. In this section, we review the studies on joint beamforming designs in RIS-assisted systems that focus on optimizing resource efficiency. These studies primarily concentrate on optimizing resource consumption by minimizing transmission power or maximizing the EE/SE of the system, with less emphasis on the overall effectiveness of the output (i.e., rate/capacity). The benefits and drawbacks of these studies are also provided. To ensure clarity in comprehending these studies, we have divided them into two groups based on their application scenarios. Tables 6–7 summarize the papers discussed below.
A. Single-Cell Multi-User Downlink MISO Scenarios
The authors of [54] propose a joint design for a multi-RIS-assisted multi-user MISO system. The considered system consists of a single BS equipped with multiple antennas that serve multiple users with the help of multiple RISs. All involved channels including the direct transmission link as well as the RISs-based channels are assumed to experience path loss and Rayleigh fading over a quasi-static block. The goal is to minimize network power consumption while ensuring that users receive high-quality service and that the RISs meet constant modulus constraints. To achieve this, the authors suggest a joint optimization design of the beamforming vectors at the BS, the active RIS set, and the corresponding phase shift matrices at the active RISs. To address the optimization problem, the authors present an AO framework. This framework optimizes the beamforming vectors of the BS, active RIS set, and phase-shift matrices at active RISs alternately. In each alternation, the beamforming optimization problem is converted to a second-order cone programming (SOCP) problem [143], while the active RIS set and the corresponding phase-shift matrices optimization problem are transformed into SDP problem via binary relaxation [144] and SDR techniques [145], [146]. The resulting SOCP and SDP problems are then solved using a CVX solver [147]. The selection of the active RISs from the entire RIS population is addressed by introducing an auxiliary binary variable a
The significance of this study lies in its potential to revolutionize the future of wireless communication by providing a more sustainable and efficient solution. This, in turn, can lead to cost savings for network operators and improved user experience by meeting QoS requirements. However, it is important to note that the study makes several simplifying assumptions, such as perfect CSI, and ignores other practical factors such as hardware impairments, which may affect the performance of the proposed algorithm. Furthermore, there are some limitations associated with the employed techniques in the proposed algorithm that should be taken into account when interpreting the results in a real-world scenario. For instance, the time and computational resources required to solve SDR problems may be substantial, particularly for large-scale RIS networks with a high number of RUs and users, which limits its practical implementation. To address these limitations, future work should explore more advanced techniques for optimization and robustness analysis. By doing so, we can ensure that the proposed algorithm is not only effective but also practical for real-world implementation.
The study in [55] presents an optimization design aimed at maximizing SE of the RIS-aided multi-user downlink MISO system by jointly optimizing power allocation at the BS and phase shift at the RIS under non-linear user proportional rate fairness constraints. The considered system consists of one BS equipped with multiple antennas, multiple users each of which is equipped with a single antenna, and RIS. The communication between the BS and the users is performed over direct link transmission as well as through the RIS. Log-normal shadowing and Rayleigh fading are considered to model the involved channels. Assuming perfect CSI and using the ZF transmission technique at the BS, the authors proposed an iterative algorithm with closed-form expressions to solve the optimization problem alternately. The proposed algorithm starts with initializing power allocation P and phase shift
The simulation outcomes have indicated that the performance of the proposed technique exceeds the aforementioned benchmark methods by a considerable margin. Moreover, the authors have evaluated the SE performance of the proposed algorithm in terms of different system parameters such as the number of RUs at the RIS and RIS location. In this context, results showed that the SE performance of the proposed algorithm is directly proportional to the number of RUs. However, it has also been observed that the SE performance is sensitive to the location in which the RIS is placed. In particular, placing the RIS in the middle between the BS and users cluster leads to the worst performance gain. However, when the RIS is placed closer to either the BS or the users’ cluster, the SE performance improves. This highlights the crucial role of optimizing the placement of the RIS in achieving maximum performance gain. The study offers valuable insights into how the number of RUs and placement can affect the performance of the proposed algorithm. However, other crucial design parameters require investigation. For example, the number of BS antennas and users are significant factors that can impact system performance and should not be overlooked. Therefore, further research is necessary to explore the influence of these parameters and their interaction with the RIS configuration.
The findings of this study hold promise for the development of mobile communication systems in the future. It highlights the need for a balanced tradeoff between SE and user fairness in RIS-aided systems. By achieving this balance, we can expect to see improved performance in mobile communication systems, which is crucial for supporting a wide range of multimedia applications. This study provides valuable insights that can guide the design and implementation of future mobile communication systems, ensuring that they meet the needs of users while also delivering optimal performance. However, it is important to be aware of the limitations of the proposed algorithm. Firstly, the algorithm assumes perfect CSI, which may not be feasible in practice due to channel estimation errors, quantization errors, and other sources of noise. This can lead to inaccurate results and reduced performance. Secondly, the algorithm is based on the ZF transmission technique at the BS, which may not be optimal in all scenarios, especially in the presence of interference. This can result in suboptimal performance and reduced efficiency. Thirdly, the algorithm may suffer from high computational complexity, especially for large-scale systems, due to the need for solving multiple optimization sub-problems in each iteration. This can result in increased processing times and reduced efficiency. Finally, the proposed algorithm can converge to a locally optimal solution and may not guarantee finding the global optimal solution due to the non-convex nature of the problem. Therefore, it is important to consider its limitations and potential drawbacks to ensure optimal performance in practical applications.
The authors in [56] have considered a multi-user RIS-assisted MISO multicast transmission system, in which a multi-antenna BS sends common messages to a group of single-antenna mobile users with the assistance of RIS. The study assumes that the BS has access to the global CSI of all links through uplink channel estimation and proposes a joint optimization design of the covariance matrix at the BS and the phase shifts at the RIS that maximizes the EE. To find a solution, the authors proposed a functional algorithm that utilizes the FP technique and the AO method. Furthermore, the authors have analyzed the EE in some asymptotic cases. That is, the order growth of the EE is obtained when the number of RUs at the RIS, the number of antennas at the BS, and the number of mobile users are infinite. Simulations are carried out to evaluate the performance of the proposed algorithm and validate the asymptotic analysis. Results showed that the proposed algorithm outperformed non-RIS multicasting systems in terms of EE by 50%. Moreover, the asymptotic behaviors of the EE show that as the number of RUs and BS antennas increases, the EE of the system increases, which means that the system can transmit more data using the same amount of energy. This is a desirable outcome as it can lead to significant cost savings and improved performance of the system. However, the study also indicates that there is an optimal number of RUs beyond which the EE starts to decrease. This means that there is a trade-off between the number of RUs and the EE of the system. Furthermore, the study shows that the EE is inversely proportional to the number of mobile users, meaning that as the number of users increases, the EE decreases. This highlights the need to balance the number of users with the number of BS antennas and RUs to achieve the highest possible EE, which is an open research direction that needs further investigation. Additionally, the study highlights the critical role of RIS placement in achieving optimal EE within the system. The results demonstrate that placing the RIS in closer proximity to either the BS or the mobile users yields higher EE than placing it in the middle of the two. These findings emphasize the significance of carefully optimizing the RIS placement to maximize the performance gain in the network.
It is worth noting that the study assumes perfect estimation of CSI, which may not be possible to achieve in practice. Moreover, the proposed optimization framework may also have computational complexity issues when dealing with large-scale systems due to its FP and AO approach. Furthermore, the proposed algorithm may only obtain local optima solutions and cannot guarantee global optimality. Therefore, further research is needed to address these limitations and improve the practicality and scalability of this proposed optimization framework. Table 6 summarizes the studies discussed in this section.
B. Other Scenarios
In this sub-section, we delve into the recent studies in joint beamforming design for optimizing resource efficiency in other considered scenarios. These studies are summarized in Table 7 to provide a clear and comprehensive understanding of the section.
The authors in [148] explored a multi-user uplink MIMO with a RIS-aided network. The considered system consists of a single RIS, one BS, and multiple mobile users. The BS and each user are equipped with multiple antennas and the uplink communication between the users and the BS is performed through RIS. The spatial channel model is considered for small-scale fading of the RIS-based channels. To this end, the authors proposed a joint optimization framework for transmitting precoding at users and phase shifts at the RIS that maximizes resource efficiency (RE), which is a performance measure that balances EE and SE. The study considers both continuous and discrete phase shifts at the RIS and partial CSI in which the instantaneous knowledge of the RIS-BS channel is perfectly known while statistical knowledge of the fast time-varying user-RIS channels is available. To solve the problem, the authors have developed an optimization framework that utilizes the AO method to iteratively update the transmit covariance matrices of the users and the phase shift values of the RIS separately and sequentially. To optimize the user’s transmit covariance matrices with fixed RIS phase shifts, the authors first derived closed-form optimal solutions for characterizing the user’s transmit signal directions. Then they proposed a simple and asymptotic SE expression and used the quadratic transformation (QT) [103] to acquire asymptotically suboptimal solutions for the user’s power allocation matrices. For the optimization of the RIS phase shift values with fixed covariance matrices, the authors handled an equivalent mean-squared error (MSE) minimization problem, which was then addressed using an inexact majorization-minimization method [149]. Numerical results demonstrated the effectiveness of the proposed optimization framework in maximizing the RE. The proposed design, in particular, significantly increased SE when compared to schemes that used equal power allocation or fixed RIS phase shifts. Furthermore, using the RIS with discrete phase shifts resulted in significant energy savings and outstanding EE gains when compared to the case with continuous phase shifts.
The same system model as [148] is investigated in [150] with both partial and perfect CSI assumptions for the RIS-based channels. The authors propose a joint optimization design for transmitting covariance matrices of all users and RIS phase shifts that maximize global energy efficiency (GEE) and SE. To tackle this problem, the authors followed the same methodology as in [148] with different techniques to design the user’s transmit covariance matrices and the phase shifts at the RIS alternately. In particular, Dinkelbach’s algorithm [151], [152] is used for solving power allocation optimization at the users while RIS phase shifts optimization has been solved based on the BCD method and the minorization-maximization technique [153]. For comparison, the authors have considered the AF relay-assisted scheme (i.e., non-RIS system) as a baseline where the RIS is replaced by the AF relay. The numerical results have demonstrated the effectiveness of the proposed optimization design in maximizing the GEE and SE as compared to the AF-relay scheme. As for the performance of the proposed design with perfect CSI, it has been shown that better GEE can be obtained as compared to the partial CSI and AF-relay scheme but with a large signaling overhead.
The results obtained from these studies have significant implications for the design and implementation of wireless systems. The proposed RIS-assisted multiuser MIMO uplink transmission scheme offers substantial benefits in terms of EE and SE, leading to enhanced network capacity and reduced power consumption. Additionally, the proposed system can handle partial CSI, which is a common occurrence in high-mobility scenarios with fast time-varying channels. This approach reduces signaling overhead while still utilizing crucial information to improve uplink transmission performance.
While the achieved results are promising, several limitations should be considered. One of the main limitations is that the proposed system assumes perfect synchronization between the users and the RIS, which may not always be feasible in practice. This assumption can limit the applicability of their proposed system to real-world scenarios. Moreover, even though the proposed system handles partial CSI, it still requires some level of channel information to operate effectively. In scenarios where channel information is limited or unavailable, alternative approaches may need to be considered. In addition, the proposed optimization frameworks have some limitations that need to be considered. One such limitation is the method employed. For example, using QT to obtain asymptotically suboptimal solutions for the user’s power allocation matrices in [148] can be computationally efficient, but it may not always provide optimal results. On top of that, although Dinkelbach’s algorithm, which was adopted in [150] for solving power allocation optimization at the users, is known to converge to the optimal solution, it may require a large number of iterations to do so. It involves solving a series of convex sub-problems which increases the computational complexity [154]. Moreover, although the use of an inexact majorization-minimization method for optimizing the RIS phase shift values in [148] and the use of the BCD method and minorization-maximization technique to optimize the RIS phase shifts in [150] can provide a good approximation to the optimal solution, their usage may not always converge to the global optimum. It is important to consider these limitations when implementing the proposed optimization frameworks. While they may not always provide the optimal solution, they can still be useful in providing a good approximation. Further research is needed to develop more efficient and effective methods for optimizing power allocation and RIS phase shifts. Finally, these studies have significant limitations as they do not consider the impact of system design parameters on the performance of the proposed optimization frameworks. These parameters include the number of RUs and placement of RISs, the number of antennas at the BS and for the users, as well as the number of users. To gain a more comprehensive understanding of the framework’s performance, further research is necessary in this area.
In [154], the authors have considered a downlink RIS-assisted MISO system consisting of one BS equipped with multiple antennas, one mobile user with a single antenna, and an active set l of the total L distributed RISs. The communication between the BS and the user is performed through direct links and the active RISs. In this work, the authors aim to maximize the EE of the network under the minimum rate requirement of the users and total power constraint. By dynamically controlling the on-off status of each RIS depending on the network requirements as well as optimizing the phase shifts of all RISs and the transmit power of the BS, the authors were able to achieve their goal. To tackle the formulated problem, the authors have proposed an AO algorithm that iteratively solves two sub-problems. As for the joint phase and power optimization sub-problem, the SCA method is used to optimize the phase shift values, and the optimal power is subsequently obtained in closed form. However, the RIS on-off optimization sub-problem is solved using the dual method and Dinkelbach’s method [151], [152]. The authors have extended their approach to multi-user scenarios in [57]. In the latter, the greedy method [155] for optimizing the RIS on-off sub-problem was proposed. The proposed iterative algorithm comprises two significant steps. The first step involves a joint optimization process, where the authors work to optimize the phase and power with a given RIS on-off vector. This step is crucial in achieving optimal performance as it ensures that the RIS system is operating at its maximum potential. In the second step, the authors update the RIS on-off vector with the optimized phase and power from the previous step. This step is essential in ensuring that the RIS system is continually optimized and performing at its best. Moreover, the authors have considered two different schemes for comparison. These are the scheme with one centralized RIS and the conventional AF relay scheme. In the latter scheme, the authors have replaced the distributed L RISs with distributed L AF relays. The performance of the considered schemes has been evaluated numerically in terms of EE. In this regard, it is demonstrated that the proposed scheme achieves up to 27% and 68% EE gains in both single and multi-user scenarios over the centralized RIS scheme and conventional AF relay scheme, respectively.
The proposed optimization schemes are significant because they can significantly improve the EE of wireless networks with distributed RISs. Improving EE is crucial for wireless systems as it can reduce their carbon footprint and operating costs while extending their battery life. Additionally, the proposed scheme can help in developing new technologies that can enhance wireless communication systems’ capabilities and provide better services to users. However, certain limitations arise in the proposed framework due to the methods employed. Firstly, the SCA method is used to optimize the phase shift values and the BS power may not always converge to the global optimum. Its performance is dependent on the initial point and the number of iterations, which can lead to suboptimal results. Secondly, the dual method and Dinkelbach’s method used to solve the RIS on-off optimization sub-problem are computationally intensive and may not scale well for larger systems. This can result in longer processing times and increased computational costs. The scalability of the framework can be improved by optimizing the algorithms used to solve the sub-problems. By doing so, the framework can be applied to larger systems without compromising its efficiency. Moreover, the study considers ideal assumptions such as perfect CSI at the transmitter and continuous phase shifts at the RIS. These assumptions could limit the generality of the achieved results in practical scenarios. Furthermore, the study does not consider the impact of hardware impairments, such as phase noise and quantization errors, on the performance of the considered system. Additionally, the study assumes that all users have the same minimum rate requirement, which may not be true in practical scenarios where users have different QoS requirements. These limitations should be taken into account when interpreting and applying the achieved results of the study to practical systems. Finally, the comparison with only two conventional schemes may not provide a comprehensive comparison with other existing schemes. Therefore, further evaluation of the proposed framework can be done by comparing it with other state-of-the-art schemes in terms of different metrics such as latency and reliability in addition to EE.
Future Research Prospects for Joint Beamforming Design
In sections III and IV, we have identified potential research gaps associated with each study. However, we will now shift our focus from individual research gaps to discussing general research directions derived from the collective findings of the discussed studies in the field of joint beamforming optimization in RIS-aided networks. Figure 6 offers a visual illustration of the future work and prospects discussed in this section.
A. Practical Limitations
As the number of research contributions on RIS-enhanced communications grows, the advantages of RISs for SE, EE, and rates are becoming increasingly apparent. However, a great number of the prevailing treatises concentrate specifically on ideal configurations such as perfect CSI and continuous phase shifts at the RIS. The current focus on these ideal cases has both positive and negative implications for the future implementation of RIS-aided networks. On the positive side, studying the ideal cases can provide useful insights into the potential benefits and limitations of RIS-aided networks, which can guide the design of practical systems. Ideal cases can serve as a benchmark for evaluating the performance of practical systems and comparing different system configurations. Moreover, ideal cases can help identify fundamental trade-offs between different system parameters and performance metrics, which can inform the development of efficient algorithms and protocols. On the negative side, ideal cases may not fully capture the practical challenges and limitations of RIS-aided networks. For example, perfect CSI may not be achievable in real-world scenarios due to channel estimation errors and feedback delays, which can degrade the performance of RIS-aided networks. Similarly, continuous phase shifts may be difficult to realize in practice due to hardware limitations and implementation complexity, which can increase the cost and energy consumption of RIS-aided networks. Therefore, to guarantee the successful implementation of RIS-aided networks, it is crucial to take into account non-ideal configurations, such as imperfect CSI and discrete phase shifts at the RIS. Additionally, several non-ideal scenarios require further investigation to support the future implementation of RIS-aided networks. These scenarios include hardware impairments, such as power amplifier nonlinearities and phase noise.
Although most previous studies have focused on ideal assumptions, few have taken into account some practical limitations. For instance, researchers in [63], [96], and [112] have examined the effects of imperfect CSI, but they have assumed ideal continuous phase shifts at the RIS. While these studies have considered the impact of channel estimation errors, it would be valuable to further investigate the robustness of the proposed algorithms to different levels of estimation errors. Developing more robust algorithms that can handle higher uncertainty levels and provide reliable performance under imperfect CSI scenarios is important. On the other hand, the studies in [58], [62], [98], [140], and [148] have considered discrete phase shifts at the RIS with perfect CSI assumption, while [52] has investigated the impact of both phase shift errors and channel estimation errors. Even though these studies have considered some practical limitations, they did not consider other constraints such as hardware impairments. Therefore, it is crucial to develop efficient designs for RIS deployment and configuration that take into account all practical constraints. Additionally, it is essential to develop efficient algorithms and techniques to overcome these limitations. This is an open research direction that requires attention and should be addressed. By doing so, we can ensure that RIS technology is optimized for real-world applications and can deliver the expected benefits.
B. Optimization Techniques
After a comprehensive review of the studies, it has become apparent that certain limitations are commonly associated with the proposed algorithms. These limitations include:
Convergence to global optimum: most of the algorithms proposed for solving non-convex optimization problems rely on FP and AO techniques, while some others adopted ML-based techniques [97], [99]. However, these algorithms only provide suboptimal solutions and may face challenges in achieving convergence to the global optimum. The presence of non-convex and non-linear fractional objective functions can result in multiple local optima, making it difficult to guarantee global optimality. To address these issues, future research could focus on developing algorithms that can better handle the non-convex nature of optimization problems and ensure convergence to the global optimum. This could involve exploring new optimization techniques or improving existing ones to provide more accurate and efficient solutions.
Computational complexity: although some of the proposed algorithms aim to reduce computational complexity compared to other existing schemes, they still involve solving multiple sub-problems and may be computationally expensive, particularly for large-scale systems. Further research could explore more efficient algorithms or hybrid techniques to reduce computational complexity while maintaining satisfactory performance. Recently, the authors in [156] suggest an interesting promising approach that combines heuristic algorithms (such as the greedy algorithm, GA, and PSO) with ML techniques. By leveraging the strengths of both approaches, this hybrid method has the potential to provide efficient and effective solutions that strike a balance between computational efficiency and solution quality in RIS-aided systems. This is a particularly important research direction that requires further investigations as RIS-aided systems are expected to scale up in the future.
Hyper-parameter tuning: an enormous number of the proposed algorithms require tuning of hyper-parameters such as step size and projection parameters. The process of selecting appropriate hyper-parameters can be challenging and may have a significant impact on the algorithm’s convergence and performance. Identifying the optimal values of these hyper-parameters is often challenging and may require manual tuning or exhaustive experimentation, which can be time-consuming and impractical. To address this issue, there is a need for research on adaptive methods or automated techniques that can effectively determine the optimal values of these hyper-parameters without extensive manual intervention.
C. Multi-Objective Optimization
The majority of existing research in joint beamforming optimization for RIS-aided networks has primarily focused on optimizing either effectiveness in the outcome of beamforming design (i.e., rate/capacity maximization) [51], [52], [53], [58], [62], [63], [96], [97], [98], [99], [112], [135], [140] or efficiency in resource consumptions (i.e., EE/SE maximization or power consumption minimization) [54], [55], [56], [57], [148], [150], [154]. This focus involves targeting one specific objective in different network configurations and assumptions. However, in multi-objective optimization, the goal is to optimize effectiveness in the outcome of beamforming design and efficiency in resource consumption simultaneously. This dual focus allows for the design of joint beamforming solutions that are not only effective but also resource-efficient. The challenge lies in defining the objective function that combines these objectives while considering their relative importance and potential trade-offs. Nevertheless, despite the potential benefits of considering multiple objectives simultaneously, only a few studies have addressed the concept of multi-objective optimization in RIS-aided networks [157], [158]. In particular, recent research [157] has explored the application of multi-objective optimization in RIS-aided multi-user MISO networks. The authors aimed to maximize the sum rate of users while simultaneously minimizing the transmit power of the BS. Another study [158] has focused on a multiuser RIS-aided SWIPT system, where the objective was to maximize both the data sum rate and the total harvested energy simultaneously. While these studies represent encouraging steps toward more comprehensive optimization solutions, further investigation and research are needed in this direction. Exploring and advancing multi-objective optimization in RIS-aided systems can potentially unlock new avenues for achieving enhanced performance and efficiency. Therefore, it is an important area for future exploration.
D. Application Scenarios
In the context of analyzing networks with multiple RISs, few studies have explored this scenario in a single-cell single-user SISO/MISO setup. For instance, the authors in [159] proposed policies for optimal RIS location-based selection, aiming to maximize the SNR at the user. Another optimization problem was discussed in [160], where the authors proposed an optimal solution for selecting suitable RISs and determining the optimal beam routing that maximizes the received signal power. However, the multi-cell multi-user (MC-MU) MIMO scenario is a crucial aspect of RIS-aided networks as it reflects the practical deployment of such networks. In this scenario, multiple RISs are deployed to cover a vast area with multiple cells and users that coexist in the same frequency band. This scenario has received less attention in the literature compared to the single-cell single-user SISO/MISO case, which is mainly due to its complexity. The coordination of multiple RISs and BSs is required to achieve optimal performance, making it a challenging scenario to study. Despite its complexity, the MC-MU case presents new opportunities for cooperation and interference management among the RISs and BSs, which can lead to significant performance gains [131]. Therefore, further research is necessary to investigate this scenario and develop efficient algorithms and protocols that can exploit the benefits of RISs in this context. Furthermore, it is imperative to explore the effects of various mobility scenarios on the efficiency of RIS-assisted wireless networks, particularly in high-speed modes of transportation such as trains and airplanes. This is a crucial area of research that requires attention. Additionally, it is essential to explore the potential of RIS for other applications beyond wireless communications, such as radar sensing, localization, and imaging. By delving into these areas, we can gain a deeper understanding of the capabilities and limitations of RIS technology and its potential for future advancements.
E. Design Parameters
Most of the studies that have been discussed analyze the impact of individual design parameters such as the number of antennas at the BS, the number of RUs at the RIS, and the number of mobile users, while very few studies considered the placement and plane position of the RIS. However, the interdependencies between these parameters and their combined effects on system performance are not thoroughly explored. Future research could focus on investigating the trade-offs and interrelationships between these parameters to provide a more comprehensive understanding of system design and develop more effective strategies for optimizing system performance.
F. Experimental Validation
Simulation studies have shown promising results for RISs, but it is essential to validate their practicality and effectiveness in real-world settings. Conducting experiments in real-world scenarios can provide more robust evidence for the effectiveness of RIS-assisted wireless networks, helping to identify any limitations and potential areas for improvement. Although there have been limited studies that have experimentally validated the feasibility and efficiency of RIS in in real-world scenarios [43], [44], [45], [46], [47], [161] there have been limited investigations into the effectiveness of RIS in large-scale, widely recognized operational commercial networks. This is particularly important in the presence of reflecting and scattering objects, with varying user densities and mobility patterns.
G. RIS Efficacy
When evaluating the efficiency of RIS-assisted systems, it is important to consider various factors beyond just power or spectral/energy efficiency. These factors include cost, complexity, and scalability, which are essential in determining the practical implementation of these systems. While RISs have the potential to improve the efficiency of wireless networks by mitigating signal attenuation, reducing interference, and increasing coverage, their practical implementation must also consider the cost and complexity of deploying and maintaining them. The scalability of RISs, i.e., the ability to increase or decrease their numbers according to network demands, is also critical for their practical implementation. Therefore, researchers and industry professionals must consider all these factors when evaluating the efficiency of RIS-assisted systems. This will help determine their feasibility and practicality in different applications and scenarios.
Performance Analysis and Parametric Studies of RIS-Aided Systems
Performance analysis and parametric studies of RIS-aided networks are essential in understanding the impact of this technology on wireless communication systems. RIS is a promising approach that has the potential to revolutionize wireless systems by improving signal quality and reducing interference. However, to fully realize its potential, it is crucial to understand the impact of various parameters on its performance. These parameters include the number of RUs, number of RIS per cell, location of RIS, number of BS antenna and user’s device antenna, as well as the characteristics of the environment and the communication system. Parametric studies offer researchers valuable insights into the optimal configuration and deployment of RISs to achieve maximum performance gains under various scenarios. These gains are measured by important metrics such as BER, OP, EC, ASEP, and system sum rate. This knowledge can pave the way for more efficient deployment of RIS technology in real-world applications.
The performance analysis techniques for RIS-aided networks differ from those used in traditional wireless communication systems. The majority of research papers that focus on evaluating the performance of RIS-aided networks utilize both analytical and simulation methods. Analytical methods are typically employed to derive closed-form expressions for performance metrics. These closed-form expressions provide valuable insights into the system performance characteristics. In addition, simulation methods are used to simulate the behavior of the network under consideration and validate the accuracy of the proposed analytical framework. However, analyzing the performance of RIS-aided systems is more complex compared to conventional wireless communication systems. This complexity arises from the introduction of RIS, which brings new challenges and complexities to the wireless propagation environment. Specifically, analytical methods for evaluating the performance of RIS-aided systems must be more sophisticated to account for the effects of RIS on the CSI and signal propagation. This sophistication is due to the dynamic nature of RIS elements, which introduce additional variables and constraints. Consequently, analytical methods for RIS-aided systems tend to be more computationally expensive than those used for traditional wireless communication systems, where the propagation environment is relatively stable and predictable. For example, the performance analysis study in [162] has attempted to model the more practically realistic large-scale, multi-cell scenarios involving multiple transmitters, receivers, and RISs sharing the same resource block (i.e., time and frequency). In such scenarios, inter-cell interference, large-scale randomness in RIS-aided channels, spatial diversity of multiple RISs, and dynamic user activities introduce significant complexity that cannot be easily simplified. To address these complexities, most performance analysis of RIS-aided wireless networks employs the stochastic geometry (SG) approach [163], which is particularly well-suited for modeling large-scale dynamic wireless networks. In contrast to traditional performance analysis techniques, which often rely on simpler and more deterministic approaches, such as ray tracing approach, the SG approach allows for a more accurate representation of the real-world complexities in RIS-aided systems. In some special cases, employing SG for performance analysis of RIS-aided wireless systems can lead to closed-form expressions that govern system behavior. These expressions facilitate the understanding of network operation and provide insightful design guidelines, which are likely to be difficult to get from computationally intensive simulations. Nevertheless, developing and implementing sophisticated and precise SG models that can effectively address the full complexity of real-world scenarios is challenging. This challenge arises from the unlimited degrees of freedom and system constraints inherent in RIS-aided wireless networks [164]. Additionally, these analytical models are often governed by tractability-accuracy tradeoffs. Precise models that accurately represent real-world complexities are often more complex and computationally demanding, posing practical implementation challenges. On the other hand, tractable models, which are simpler and easier to solve and analyze, often sacrifice accuracy in modeling realistic scenarios, thereby limiting the insights they can provide. One potential solution to address these challenges is to use a combination of SG and ML techniques [164]. This approach can help to overcome this tradeoff by leveraging the strengths of both techniques. However, using ML techniques requires large amounts of data to train the models. In some cases, data may be limited or difficult to obtain, which can make it challenging to develop accurate models. Thus, a careful balance between model sophistication and practical feasibility is crucial for effective use in real-world RIS deployments. Furthermore, validating the analytical models in real-world scenarios is a critical step in ensuring their accuracy and reliability. This can be achieved through a combination of extensive field trials and simulations. In the field trials, RIS prototypes are deployed in diverse environments, whereas simulations are conducted to model the expected behavior of RIS in similar scenarios. If the field trial results closely match the simulated results, then it provides strong evidence that the analytical models are accurate, reliable, and reflect real-world performance. For example, the authors in [46] conducted field trials of RIS prototypes in various real-world 5G networks, including indoor, outdoor, and mixed environments. Measurement results from these trials were closely aligned with the simulation results, thus validating the accuracy of the analytical model and confirming the effectiveness of RIS in enhancing 5G network coverage.
The analytical and simulation methods used in the performance analysis of RIS-aided networks are adaptable and can be designed to handle dynamic environments and rapidly changing user requirements. Analytical methods such as SG can incorporate complex models that account for uncontrollable dynamic changes in the wireless environment. Using random processes, these models may involve equations that consider uncontrollable variations in signal propagation, channel state information, and interference patterns. Researchers can parameterize their analytical models to include uncontrollable variables representing changing user requirements and environmental conditions. This allows them to evaluate how RIS-aided networks perform under dynamic scenarios. On the other hand, simulation-based methods are highly flexible and can simulate dynamic scenarios by introducing variables that represent changing user demands and environmental factors. Researchers can adjust these variables to simulate different conditions. However, the computational complexities associated with these methods in RIS-aided networks are relatively high. Analytical methods involve solving mathematical equations that model the behavior of RIS in the wireless environment. These equations can be complex and computationally intensive to solve, particularly when considering the dynamic nature of RIS interactions with signals and channels. Simulation methods, on the other hand, require creating detailed computer models of RIS-aided systems. These models need to account for numerous variables, such as signal paths, interference patterns, and RIS configurations, making the simulations very complex. Running these simulations to accurately reflect real-world scenarios demands significant computational resources, leading to high computational complexity. Researchers need to strike a balance between model complexity and computational efficiency to achieve meaningful results. Additionally, the availability of real-world data for calibration and validation plays a crucial role in enhancing the accuracy of these methods in dynamic scenarios.
One valuable way that has been widely adopted to measure the effectiveness of the analytical methods in the performance analysis of RIS-aided networks is to compare the analytical results with simulation results. This comparative approach serves as a benchmark to evaluate the accuracy of the analytical models. If the analytical results closely align with the simulation results, it indicates that the analytical method is effective in capturing the complex interactions within RIS-aided networks. This validation process builds confidence in the analytical models and demonstrates their utility in providing insights into network performance. Furthermore, analytical models often employ approximations, and the validity of these approximations can vary with the scale of the system. For instance, the central limit theorem (CLT), which is widely adopted in the literature is applicable primarily in scenarios with a large number of RUs and in low SNR regimes. In contrast, gamma approximations offer greater flexibility, making them suitable for analyzing systems with varying and arbitrary numbers of RUs. This scalability factor is significant when assessing the effectiveness of analytical methods in modeling RIS-aided networks, as it demonstrates their adaptability to different deployment scenarios and network sizes.
However, environmental factors, including physical obstructions and atmospheric conditions, can impact the effectiveness of RIS performance analysis in several ways. Firstly, physical obstructions like buildings and obstacles can obstruct the LoS between RIS elements, transmitters, and receivers, leading to complex signal propagation with multipath effects. Analyzing how these obstructions affect signal strength and quality requires precise modeling and simulations. Secondly, atmospheric conditions like rain, fog, and humidity can attenuate radio signals, affecting received signal power, SNR, and network performance. RIS performance analysis must consider these atmospheric effects, especially in outdoor scenarios. Thirdly, physical obstructions can create interference by reflecting signals unpredictably, particularly in dense urban environments. This interference complicates the analysis of RIS-aided networks. Additionally, the presence of physical objects can lead to extra path loss due to signal absorption and scattering, making accurate path loss models crucial for RIS performance analysis. Finally, dynamic environmental changes, including weather fluctuations and moving objects, add complexity, requiring RIS performance analysis to adapt to evolving conditions. To address these challenges, RIS performance analysis should employ analytical and simulation techniques that realistically model environmental factors, ensuring reliable assessments of signal propagation, SNR, and other critical metrics. Real-world measurements and field trials are also necessary to validate models under varying environmental conditions.
Despite theoretical advancements, implementing RIS performance analysis in real-world scenarios demands specific hardware and infrastructure requirements. For analytical methods, powerful computational resources are necessary due to the complexity of solving complex mathematical equations. High-performance servers or computing clusters equipped with advanced processors and sufficient memory are essential for the efficient execution of analytical models. Additionally, simulation methods require specialized software and hardware setups. High-performance computers with multi-core processors are essential for running detailed simulations. Furthermore, storage capabilities are necessary to store large datasets generated during simulations. Specialized simulation software that can accurately model the behavior of RIS-aided networks is important. In real-world scenarios, network emulation tools and testbeds are often employed to validate analytical and simulation results. These tools require physical setups with RIS devices, antennas, and other network components. The infrastructure must support precise control over signal paths, interference patterns, and RIS configurations, allowing researchers to conduct experiments that mirror real-world conditions accurately. These requirements enable researchers to conduct in-depth analyses, validate theoretical models, and optimize RIS configurations for practical deployment in wireless communication systems.
In this section, we will explore recent developments and challenges in the performance analysis and parametric studies of RIS-aided systems. To provide a thorough understanding of the latest developments in this field, we have organized the relevant literature into two main distinct categories, primarily based on the number of RIS used. The purpose of this categorization is to enhance clarity and enable a systematic exploration of the subject matter.
A. Single RIS-Assisted Wireless Systems
In this section, we delve into the performance analysis and parametric studies of wireless networks enhanced by the utilization of a single RIS. To provide a comprehensive understanding of this scenario, we have thoughtfully divided this section into two subsections, based on the specific application scenarios considered. These studies are summarized in Tables 8–10.
1) RIS-Assisted Wireless Networks of Single-Antenna Transceivers’
Within this subsection, we further subdivided the scenario into two additional sub-sections, based on the user’s mobility. This division allows for a more detailed examination of the subject matter and its implications.
a: RIS-Assisted Wireless Networks of Static Single-Antenna Transceivers’
The authors in [165] have considered a point-to-point downlink SISO with an RIS-aided system where a single antenna source communicates with a single antenna destination via an RIS. The direct link transmission between the source and the destination is assumed to be blocked while the Rayleigh fading channel is adopted to model RIS-based channels. The study assumes optimal phase shifting at the RIS where the phases of the channels are perfectly known to the RIS. To this end, a framework for assessing the EC has been reported. In this context, closed-form expressions for the probability density function (PDF) of the end-to-end (e2e) fading channel coefficient are presented for multiple and single RUs. Based on them, closed-form expressions for the EC have been extracted for both cases (i.e., for single and multiple RUs). In addition, the EC is asymptotically analyzed for high SNR and a high number of RUs where tight approximations for the EC are reported. Monte Carlo simulation has been conducted to emphasize the precision of the proposed framework. Results demonstrated that increasing either the number of RUs or the transmission SNR leads to a proportional improvement in the EC. For instance, when considering an RIS with 2 RUs, increasing the transmission SNR from 5 to 10 dB yields a significant 34% gain in the EC. Likewise, for a fixed transmission SNR of 10 dB, increasing the number of RUs from 50 to 100 results in a notable 13% gain in the EC. This suggests that deploying a larger number of RUs or increasing the transmission SNR levels results in improved EC performance, allowing for increased data rates. This highlights that RIS technology can be particularly beneficial in scenarios with challenging SNR conditions, such as long-range or interference-prone environments. Additionally, the results demonstrate that the achieved gains in EC are scalable. Doubling the number of RUs consistently leads to approximately 2 bits/s/Hz improvement in the EC, regardless of the SNR level. This scalability implies that as RIS technology advances and larger arrays of RUs become feasible, further performance improvements can be expected. It is important to note, however, that the study does not explicitly consider the placement of the RIS. The performance of RIS-aided systems can be influenced by this factor as it plays a crucial role in determining the effectiveness of RIS-based signal manipulation, and their impact should be further investigated. In addition, while the study offers closed-form expressions and asymptotic analysis for the EC, it overlooks other crucial performance metrics such as SE and EE. These metrics provide invaluable insights into the overall system performance and efficiency, and thus, warrant further investigation.
Although the study offers valuable insights into RIS-aided systems and their performance, it is important to consider the limitations associated with the system model and assumptions used. The study focuses on a specific system configuration with a single-antenna source, one single-antenna destination, and an RIS. However, real-world wireless networks are much more complex and may involve multiple BSs, multiple mobile users, and interference from other sources. Therefore, the study findings may not directly apply to such multi-cell or interference-limited scenarios. Furthermore, the study assumes a blocked direct link transmission and a simplified channel model for RIS-based links where independent and Rayleigh-distributed channels are assumed. However, wireless channels are affected by various factors such as path loss, shadowing, and multi-path propagation. Ignoring these factors may not accurately capture real-world channel behavior. Additionally, the study overlooks the potential impact of LOS links and other deterministic components by neglecting the deterministic path gain in the fading coefficients. In practice, these components can significantly affect the system performance, especially in scenarios with strong LOS or obstructed links. Moreover, the assumption that the RIS has perfect knowledge of the channel phases may not be feasible in practical scenarios. In reality, CSI estimation is subject to estimation errors and imperfections. Ignoring the estimation process may lead to an overoptimistic assessment of system performance. Additionally, assuming equal reflected gain for all RUs simplifies the analysis but may not reflect the actual behavior of practical RISs. In reality, the reflected gain may vary across different RUs due to hardware imperfections, varying distances, and orientations. Therefore, considering unequal reflected gains would provide a more realistic representation of RIS performance. Finally, the results highlight the benefits of increasing the number of RUs or the transmission SNR, but they do not consider potential trade-offs or practical constraints associated with scaling up the RIS or increasing the SNR levels. Therefore, further research is needed to address these limitations and explore the implications of the findings in more complex and realistic scenarios.
The authors in [166] have considered a downlink SISO wireless communication system aided by RIS. The system involved a single-antenna source communicating with a single-antenna destination via an RIS, with the direct link transmission between the two being blocked. The study assumes full knowledge availability of CSI for all involved channels. To model the system, the authors employed a realistic path loss model that took into account the physical size of the RIS and the effective angle of incidence as a large-scale channel model and a Rayleigh fading model as a small-scale channel model. The performance of the system is evaluated in terms of SNR coverage probability for a single and arbitrary number of RUs at the RIS. More precisely, for the single RIS element case, the authors obtained an exact closed-form expression for SNR coverage probability by determining the distribution of the corresponding channel gain under Rayleigh fading based on Meijer’s G-function [167]. For the multiple RUs case, the gamma distribution is adopted to approximate the distribution of the resultant channel gain by the moment matching method [168], [169]. Upon this, a general and accurate expression for SNR coverage probability for an arbitrary number of RUs is obtained. The authors also quantified and assessed the effects of critical factors on the system coverage performance, such as the number of RUs, physical RIS size, fading channel coefficients, the effective angle of incidence, and the placement of the RIS plane. It is important to note that the placement of the RIS plane has been assessed based on the effective incidence angle,
![FIGURE 7. - Illustration of RIS rotation and corresponding angles considered in [166].](/mediastore/IEEE/content/media/6287639/10005208/10356072/mahmu7-3342320-large.gif)
The study’s findings contribute to the practical implementation of RIS technology in wireless communication systems. By providing insights into critical system factors and their impact on performance, the results help guide the design and deployment of RIS-assisted systems in real-world scenarios. Nevertheless, it is worth mentioning that the expressions derived in the analysis are based on Gamma approximation. This approximation is deemed inadequate for diversity analysis as it is unable to fully extract the diversity order [170]. Moreover, it is important to note that there are several potential limitations associated with the study. For instance, the study has made certain simplifying assumptions in its system model, such as considering a single antenna source, and stationary single antenna destination, assuming complete knowledge of CSI, ignoring hardware imperfections and interference from other devices, and considering ideal continuous phase shifts at the RIS. These factors could potentially impact the performance of RIS-assisted systems differently in real-world scenarios. These simplifications may not fully capture the complexity and diversity of real-world communication scenarios, limiting the generalizability of the findings. Furthermore, the study assumes blocked direct links and a simplified channel model for RIS-based links, which may not fully represent real-world propagation characteristics. In reality, wireless channels are affected by various factors such as shadowing, and multipath fading, which are not explicitly considered in the model. Additionally, the study assumes an ideal reflection coefficient of one for all RIS RUs, indicating perfect reflection. In practice, the reflection coefficients may vary due to imperfections in the RIS elements, environmental factors, or calibration errors. Ignoring these variations may lead to optimistic performance estimates that may not hold in real-world implementations. Finally, the study assumes far-field transmission which may limit the applicability of the findings to scenarios that involve short-range communication (i.e., near-field transmission). Therefore, it is crucial to conduct further research to address these limitations and ensure the practical implementation of RIS technology in wireless communication systems. By doing so, we can unlock the full potential of RIS-assisted systems and revolutionize the way we communicate wirelessly.
The authors in [171] have considered a downlink RIS-aided SISO communication system and emphasized the impact of transmission SNR, SNR threshold
The system under investigation involved a single antenna source that communicates with a single antenna destination via RIS. The direct link between the source and destination is assumed to be blocked by an obstacle, and the RIS-based channels are modeled using the Rayleigh fading model. The authors assumed perfect CSI of the channels at the RIS to enable the realization of optimal phase shifting. To this end, the authors have derived a novel closed-form expression for the e2e signal-to-noise-plus-distortion-ratio (SNDR) and an exact closed-form expression for the system OP. Additionally, they have reported a new upper bound for the EC in terms of the aforementioned system parameters. The findings reveal that for a fixed level of hardware imperfections, as the transmission SNR increases, the system OP performance improves. Moreover, higher RUs values lead to higher diversity gains and better outage performance. The results also show a trade-off between system spectral efficiency and power consumption. For a fixed number of RUs and a given OP requirement, increasing spectral efficiency (i.e., increasing
The significance of this study lies in its comprehensive analysis of the impact of hardware imperfections on RIS-assisted wireless systems. The results obtained can be used to design and assess RIS-aided systems in practical scenarios where hardware imperfections are inevitable. Although the study comprehensively examines the joint effects of the parameters under investigation and explores their combined impact on system performance, it is important to note that additional design parameters, such as RIS placement and user mobility, should not be disregarded. Further research in this direction is essential to gain a more comprehensive understanding of the optimal RIS placement and the influence of user mobility on system performance. In addition, the study has some limitations due to assumptions made that need to be addressed. For instance, the study assumes that the direct link transmission was blocked while the RIS-based channels were assumed to be independent and identical fading channels, following a Rayleigh distribution. However, this simplistic channel model neglects the presence of more complex channel characteristics such as path loss, shadowing, and multipath propagation, which can have a significant impact on the performance of the RIS-aided network. Moreover, the study models hardware imperfections at the transmitter and receiver as zero-mean complex Gaussian processes with specific variance parameters. This idealized representation may not capture the true nature and extent of hardware imperfections that can exist in practical systems, leading to a gap between the model and real-world performance. Additionally, the study neglects other sources of impairment such as interference from other users or neighboring cells, phase noise, and nonlinearities in the transceiver components, which can also affect the performance of these systems. To address these limitations, future research should investigate the joint effects of hardware imperfections and other sources of impairment. Furthermore, the study assumes ideal continuous phase shifts at the RIS and that the RIS has perfect knowledge of the phase of the channel and applies an optimal phase shift for each RU based on this knowledge. However, in practical scenarios, a limited number of phase shifts are available at the RIS, perfect knowledge is challenging to achieve, and the optimal phase shift may be challenging to determine due to channel variations, estimation errors, and limited feedback capacity. These assumptions may overstate the performance of the RIS-aided system. Finally, the study assumes that the RIS response magnitude is fixed at one for simplicity. However, in reality, the response magnitude may vary depending on various factors, including element characteristics, frequency, and incident angle. Neglecting these variations can result in an inaccurate representation of the RIS’s impact on system performance. By addressing these limitations, future research can provide a more accurate and comprehensive understanding of the potential benefits and challenges of RIS-aided networks in practical scenarios.
The authors in [172] conducted a quantitative analysis of the performance of RIS-aided downlink SISO communication systems. The study focused on coverage, probability of SNR gain (PSG), and delay outage rate (DOR). The system under consideration involved a single antenna source communicating with a single antenna destination via RIS. The direct link between the source and destination is assumed to be blocked by an obstacle, and the RIS-based channels are modeled using a quasi-static flat Rayleigh fading model. The authors assumed perfect CSI of the channels at the RIS to enable the realization of phase shifts. To provide a basis for comparison, the authors analyzed other transmission schemes, such as dual-hop relaying systems and direct transmission systems. The PSG is used to represent the probability that the received SNR at the destination is higher when utilizing RIS compared to relayed or direct transmission systems. On the other hand, the DOR served as a valuable performance metric for designing URLLC transmission systems. It quantified the probability that the duration needed to successfully transmit a specific data quantity in a wireless channel exceeded a predefined threshold.
The authors derived mathematical expressions for the aforementioned metrics based on CLT approximation [173], and simulations were performed to validate the correctness of the derived expressions. The study also investigated the impact of practical discrete phase shifts on PSG performance. The findings showed that practical discrete phase shifts resulted in a significantly lower PSG performance than the ideal phase shifts. However, this loss could be mitigated by reducing the interval of discrete phase shifts (i.e., increasing the number of quantization levels). Furthermore, the study found that the number of RUs and transmission SNR has a significant positive impact on the coverage performance, and the adoption of RIS could enhance the coverage distance, boost SNR gain, and minimize DOR compared to the benchmark systems.
The findings of this study have significant implications for the design and optimization of wireless communication systems, as they suggest that RISs can effectively address limitations inherent in traditional relaying systems and direct-link transmission. Additionally, the study provides valuable insights into the potential applications of RISs in various scenarios, such as URLL transmissions. Nevertheless, it is important to note that the expressions derived in the analysis are based on the CLT approximation. This approximation guarantees accurate results when there is a large number of RUs. However, it becomes less reliable when the number of RUs is small [170]. Furthermore, the CLT approximation may exhibit poor performance in high SNR scenarios [174]. Therefore, the applicability of the achieved results might not be suitable for scenarios with a small number of RUs or high SNR. Therefore, future research should investigate more sophisticated and precise analytical models to assess the performance of the network under high SNR conditions. Moreover, the study has several limitations due to the assumptions made. Firstly, it assumes a simplified quasi-static flat Rayleigh fading channel without considering more realistic fading conditions such as frequency-selective and correlated fading. This oversimplification may not accurately reflect real-world channel characteristics. Additionally, the study only accounts for the signal reflected by the RIS the first time, neglecting subsequent reflections that could potentially impact system performance. The absence of considering user mobility is another limitation, as it restricts the generalizability of the findings to dynamic scenarios where users are moving and experiencing changing channel conditions. User mobility introduces additional challenges that can affect RIS-aided system performance. Furthermore, the study assumes a simplified far-field path-loss model based on a specific distance criterion, which may not fully capture the complexity of path-loss behavior, especially in scenarios where the distances between the RIS, source, and destination are closer or fall within the near-field region. Additionally, the assumption of ideal phase shifts for the reflection coefficients produced by the RIS may not reflect the practical limitations of RISs. Achieving perfect phase control and maintaining continuous phase shifts can be challenging in practice. Moreover, the study assumes that the RIS has perfect knowledge of the channel phases, which may not align with the practical limitations of CSI acquisition. Acquiring accurate and up-to-date CSI introduces challenges, including overhead, estimation errors, and feedback limitations. Ignoring these practical limitations of CSI acquisition may lead to an overestimation of system performance. These limitations should be carefully considered when interpreting and applying the study’s results in practical RIS-aided communication systems. Consequently, further research is required to validate these findings and explore their applicability in diverse scenarios. Additionally, the study primarily focuses on a subset of design parameters, such as the number of RUs and transmission SNR. However, other crucial factors, including RIS placement, user mobility, interference conditions, and varying network topologies, need to be considered. Future research should investigate the interplay and impact of these design parameters to achieve a comprehensive understanding of RIS-aided systems’ performance and their suitability in different deployment scenarios.
The authors in [175] evaluated the outage and error performance of a SISO wireless system aided by RIS over a Nakagami-
Moreover, the authors have explored a more practical RIS configuration, where a quantized phase shift scheme is employed. In this context, and in the absence of a direct link, the authors have studied the EC of the OPS, 2-bit quantized PS, and RPS schemes as a function of the distance between the source and the RIS as well as the number of RUs. The results have shown that in all designs, a minimum value of EC is observed when the source-RIS distance is approximately 50 meters. Moreover, the quantized PS scheme exhibits superior performance compared to the RPS design across all values of the number of RUs in the RIS. However, it has been noted that there is a slight decrease in performance when utilizing the quantized scheme in comparison to the OPS design. Specifically, a loss of approximately 0.6 bps/Hz in EC is observed when using RIS with 320 RUs. This finding suggests that practical implementations of RIS technology can achieve performance gains similar to those of ideal RIS configurations with minimal additional hardware complexity. Therefore, developing new algorithms and techniques for optimizing phase configurations that strike a better balance between performance and complexity is an interesting open research direction that requires further investigation.
Although the study provides valuable insights into the design of RIS-empowered systems for optimal performance under realistic conditions, several limitations must be considered. One of the limitations is that the study assumes flat fading channels, where the fading coefficients remain constant over the entire bandwidth. This assumption neglects the effects of frequency-selective fading, which can have a significant impact on the system’s performance in real-world scenarios. Additionally, the study assumes that the fading coefficients for the direct and RIS-based channels are independent and identically distributed (i.i.d.). However, in practice, these coefficients may exhibit correlation and variation, which can affect the system’s performance. Ignoring these factors may lead to unrealistic performance estimations. Moreover, the study assumes that the phases of the fading coefficients and RIS element responses are uniformly distributed between 0 and
b: RIS-Assisted Wireless Networks of Mobile Single-Antenna Transceivers’
The authors in [64] have evaluated the achievable data rate performance of an uplink RIS–aided SISO communication system with limited phase shifts at the RIS. The considered system consists of one BS equipped with a single antenna, one user with a single antenna, and RIS. The direct link transmission between the user and the BS is assumed to be blocked and Rician fading is adopted to model the channel between BS and the user via RIS. To this end, the authors derive an approximation of the achievable data rate with continuous phase shifts of the RIS. This approximation involves determining both upper and lower bounds of the data rate, based on the channel model and propagation characteristics. The upper bound is reached when the Rician factor approaches infinity (i.e. when the channel is pure LoS), while the lower bound is achieved when the Rician factor approaches zero (i.e. when the channel is Rayleigh). They then explored how limited phase shifts impact the data rate based on the derived expression. The authors also quantify the necessary number of phase shifts that are needed to ascertain that the performance degradation remains at a level below the predetermined limit. Moreover, the study investigates the impact of the number of RUs as well as the RIS placement on the achievable rate performance. Simulations are carried out and validate the correctness of the theoretical framework. The results imply that the number of RUs has a positive correlation with data rate performance. In addition, for discrete phase shifts, as the number of RUs increases, the required number of phase shifts decreases. For instance, a RIS with three RUs requires three coding bits, while a RIS with infinite RUs requires one coding bit. These results highlight the importance of considering the number of RUs when implementing discrete phase shifts, which can lead to more efficient and cost-effective implementation. Furthermore, the performance degradation of the data rate is analyzed in relation to the distance between the BS and the RIS. The findings reveal that for an RIS with three RUs and two coding bits, as the distance increases, the data rate degradation initially decreases and then increases. The optimal distance that minimizes degradation and maximizes data rate is found to be 80 meters. This behavior is attributed to the fact that the channel gain follows a similar pattern. The performance degradation is directly influenced by the channel gain, which explains the observed trend. Additionally, the magnitude of variability caused by changing the location of the RIS is influenced by the number of the RUs and the number of coding bits. A higher number of RUs or higher numbers of coding bits result in smaller variations in performance degradation when the RIS location changes. For example, when using three RUs and three coding bits or using 300 RUs and two coding bits, the performance degradation does not vary significantly despite changing RIS locations. This implies that increasing the number of RUs or the number of coding bits can help mitigate the impact of RIS location variations on performance degradation.
The importance of this study lies in its potential to aid in the development of practical RIS-assisted systems that have limited phase shifts. However, the study has several limitations that should be considered when interpreting its results. Firstly, the scenario assumed in the study is simplified, focusing on a narrow-band uplink cellular network with only one BS and one user. This oversimplification may not accurately represent the complexities of real-world cellular networks with multiple BSs, users, and interference. Secondly, the study assumes that the LoS link between the user and the BS may be unstable or experience a complete outage, neglecting other factors like multi-path propagation, shadowing, and interference that are commonly present in real-world scenarios. Thirdly, the study assumes a reflection-dominant channel model, overlooking the potential impact of weak reflected signals or significant NLoS components that can exist in real-world channels. Fourthly, the study assumes ideal RUs in the RIS, disregarding practical limitations such as imperfect phase control, hardware impairments, and finite channel estimation accuracy. Additionally, the study assumes fixed reflection factors for each RIS element, neglecting variations that can occur due to element location, signal frequency, and incident angle. Furthermore, the study provides an analysis of the achievable data rate based on the assumptions and system model. However, the analysis overlooked important factors that can impact the data rate, such as higher-order modulation schemes, channel coding, resource allocation, and interference management. These factors are crucial in practical cellular networks and can significantly affect the actual data rate performance. It is important to acknowledge these limitations, and further analysis and real-world experimentation are necessary to validate the findings of the study. Additionally, the study highlights the importance of considering the number of RUs and placement of the RIS when implementing discrete phase shifts for efficient and cost-effective implementation. Future research should explore cost-effective approaches for RIS implementation, such as low-cost RUs or efficient phase shifters. This could involve designing novel RIS architectures and practical implementation techniques that balance performance gains with cost considerations.
In [65], a multi-user SISO with a RIS-aided wireless system has been investigated where a single-antenna BS serves multiple single-antenna users with the assistance of an RIS. The authors have considered two cases for the direct link transmission between the BS and users; the first case assumes that there exists a direct transmission path that experiences a Rayleigh fading while the second case assumes that the direct transmission is blocked. Moreover, the BS-RIS and RIS-users channels are assumed to follow the Rician fading model. Furthermore, the authors have considered two separate scenarios depending on the CSI availability at the RIS, namely without CSI and with CSI. For the without-CSI scenario, an RPS scheme is adopted, while an OPS scheme is used for the with-CSI scenario. The main focus of the study is on evaluating the system performance in terms of EC for both scenarios, taking into account multi-user scheduling. In this context, the authors consider a scenario in which only one user is selected for transmission at any given moment. In this scenario, they assume the presence of feedback links between the BS and all the users. The feedback links inform the BS about the channel gain, thus allowing the BS to choose and schedule only one user with the highest channel gain to transmit at any particular timeslot. To this end, the authors derive closed-form expressions for the EC based on CLT approximation and perform asymptotic analysis to investigate the impact of key system parameters such as the number of users (M), the number of RUs (N
The implications of the achieved results provide valuable insights for the design, optimization, and operation of RIS-aided multi-user wireless communication systems. They inform system designers and researchers about the potential benefits and performance characteristics of RIS technology, guiding them in making informed decisions regarding system parameters, CSI acquisition strategies, user scheduling, and LoS considerations. Nevertheless, it is crucial to acknowledge that the derived expressions in the analysis rely on the CLT approximation. While the CLT approximation ensures accuracy when there is a large number of RUs, its reliability diminishes as the number of RUs decreases [170]. Additionally, the CLT approximation may yield unsatisfactory results in high SNR scenarios [174]. Therefore, the obtained results may not be suitable for scenarios with a small number of RUs or high SNR. Consequently, it is essential for future research to explore more advanced and precise analytical models that can effectively evaluate network performance under high SNR conditions. Moreover, the study considers two scenarios based on the availability of CSI at the RIS: without CSI and with CSI. However, it does not explore the impact of imperfect or partial CSI, which is more realistic in practical scenarios. Investigating the performance of RIS-aided systems with imperfect CSI could provide insights into the robustness of the system and the trade-offs between performance and CSI acquisition complexity. Furthermore, the study assumes single-antenna deployment at the BS and users, which limits the applicability of the achieved results to the considered scenarios. In practice, multi-antenna transceivers are common, and their interactions with the RIS could affect system performance. Considering MIMO-based systems and analyzing their impact on system performance would provide a more realistic assessment. Also, it is worth noting that the study focused on evaluating the effects of parameters such as direct link transmission, CSI availability, number of users, number of RUs, and Rician fading factor. However, the study does not consider other important parameters such as RIS plane position and placements, which could significantly influence the system’s performance. Future research in this direction is needed to investigate the individual and combined effects of these parameters and their optimization. Additionally, there are several limitations associated with the study due to its assumptions. For instance, the study assumes i.i.d. complex Gaussian random variables for the LoS and NLoS components of the channel. However, this simplification may not accurately capture the real-world propagation characteristics, which can exhibit more complex behavior such as spatial correlation and multipath fading. Furthermore, the study assumes a uniform planner array configuration for the RIS, which simplifies the analysis but may not represent the actual RIS implementations in real-world scenarios. Different RIS geometries or non-uniform element spacing can impact the performance and introduce additional complexities. Moreover, the study assumes equal large-scale path-loss coefficients for all users, overlooking the potential variations in the channel conditions experienced by different users. In reality, users can be located at different distances from the BS or experience different environmental conditions, resulting in varying path losses. Additionally, the study assumes that only a single user is selected for transmission at each instant. However, in practical scenarios with multiple users, the interference between users and the impact of user scheduling policies can significantly affect system performance. Disregarding these factors may lead to an inaccurate representation of the system performance under practical conditions. Therefore, it is essential to consider the complexities introduced by multiple users and incorporate realistic user scheduling policies to obtain a more comprehensive understanding of the system behavior in real-world deployments.
The authors in [178] provide a comprehensive performance analysis of a RIS-assisted SISO system operating over Rayleigh fading channels. This system relies solely on communication through the RIS between a single-antenna BS and a single-antenna user and assumes complete knowledge of CSI at the BS. The authors consider the closed-form empirical model that takes into account the relationship between the amplitude and phase response of the RUs of the RIS, as established in [179]. To this end, the authors derive closed-form expressions for the OP and EC of the system based on multivariable Fox’s H-function [180]. To gain a further understanding of how system parameters affect performance, they also develop asymptotic expressions for these metrics under the assumption of a large number of RUs using CLT. Additionally, the authors establish both upper and lower bounds for the EC using Jensen’s inequality [101] and validate their derived expressions through Monte Carlo simulations. Based on their results, the authors quantify the effect of different system parameters on OP and EC performance. These parameters include transmitting SNR, number of RUs, minimum values of the reflection amplitude
The study’s consideration of practical aspects, such as the relationship between the amplitude and phase response of the RUs, contributes to the practical implementation of RIS technology. It bridges the gap between theoretical analysis and real-world deployment, ensuring that the achieved results are applicable and relevant in practical RIS-assisted systems. Nevertheless, it is crucial to consider other design parameters, including the placement of the RIS, and analyze their interaction with other parameters to assess their impact on the system’s performance. Further investigation in this area is necessary for future research. Moreover, there are several avenues for future research that the implications of the achieved results highlight. These areas call for further exploration and investigation to enhance the understanding and utilization of RIS technology in wireless communication systems. One crucial direction for future research is the consideration of imperfect CSI, discrete phase shifts at the RIS, and extending the proposed approach to more practical system models that involve multiple antenna transceivers and multiple users in the presence of direct links. These scenarios represent more realistic wireless communication setups, and therefore analyzing the performance of RIS-assisted systems in such settings would contribute to their practical adoption. Additionally, there are several limitations due to modeling assumptions that the authors made. For instance, the study assumes that the signals transmitted from the BS are appropriately reflected by the RIS by dynamically adjusting the phase shifts of the RUs. It ignores the reflections that occur multiple times between the RIS and the user. In reality, multiple reflections can significantly affect the channel characteristics and system performance. It is important to consider these limitations when interpreting the results of the study and applying them to real-world RIS-assisted communication systems. Further research is necessary to validate the findings in more realistic settings and under different conditions.
The authors in [181] evaluated the coverage performance of a downlink RIS-assisted SISO system over the Nakagami-
The study offers valuable insights into how RIS can enhance wireless network coverage and presents a comprehensive framework for analyzing their performance under different operating conditions. However, it is important to consider some limitations. Firstly, the proposed framework assumes a generic Nakagami-
In [182], the authors have evaluated the error performance of the RIS-aided downlink SISO wireless network in the presence of direct links and with discrete phase shifts at the RIS. The considered system involves a single-antenna transmitter that serves single-antenna mobile user with the assistance of RIS. The Rayleigh fading model is utilized to model both the direct links and the RIS-receiver links, while the Rician fading model is adopted for the transmitter-RIS LoS link. To this end, the authors have derived exact and asymptotic BER expressions considering binary phase shift keying (BPSK) modulation scheme using CHF based approach, Gauss Chebyshev Quadrature (GCQ) rule [183], and Fox’s H-function. Monte Carlo simulations were carried out and validate the correctness of the theoretical framework. In particular, the results demonstrated a strong concurrence between the analytical expression for the BER obtained as well as its corresponding high SNR approximation and the results obtained from simulations. Moreover, findings have illustrated that an increased number of quantization levels results in lower BER and that performance gain is obtained when increasing the transmission power. These findings suggest the importance of higher resolution in phase control for improved performance and indicate the significance of power allocation strategies in RIS-aided systems. Furthermore, the authors have highlighted the user operations mode with and without direct link by studying the impact of the RIS-user distance on error performance for different numbers of RUs. They find that in the absence of a direct link, the user can choose to operate solely via the RIS if the RIS-user distance is within certain thresholds, which are specified for different numbers of RUs. For instance, the user can choose to operate via RIS if the RIS-user distance is less than 50 meters for 30 RUs and less than 30 meters for 10 RUs. This provides practical guidance for determining the optimal operating mode in RIS-aided systems based on the RIS-user distance. However, if a direct link is available, it is advantageous to combine both signals, especially as the number of RU’s increases. This suggests that the cooperative use of both paths can enhance system performance and maximize the benefits of RIS technology.
The achieved results contribute to the understanding of error performance in RIS-aided wireless networks and provide valuable insights for system design, parameter optimization, and operational decision-making. Nevertheless, it is crucial to investigate the impact of other design parameters such as the RIS plane positioning and its placement from the transmitter, along with the considered parameters. This will enable the identification of optimal configurations and strategies to maximize error performance, thereby facilitating the development of highly efficient RIS-aided systems. Moreover, it is important to note that the proposed framework assumes that the RIS is perfectly reflective and does not account for the impact of losses or other imperfections in the RIS. In reality, RISs may have non-ideal reflection coefficients due to manufacturing imperfections or environmental factors, which can affect the system performance. Therefore, further investigation is needed to understand the impact of non-ideal RIS reflection coefficients on system performance. This can involve developing new analytical frameworks or simulation models that consider the effects of losses and other imperfections in the RIS. Additionally, the study assumes uniform quantization of channel phase shifts with a fixed number of levels. However, in practice, the number of quantization levels may vary depending on implementation and hardware constraints, which can also affect system performance. Therefore, exploring different quantization schemes, such as non-uniform or adaptive quantization schemes, that adjust to changing channel conditions is an open research direction. Furthermore, while the study considers practical discrete phase shifts at the RIS, other practical implementation issues such as hardware constraints and imperfect CSI are not considered. It is important to address these issues to fully understand the practical utilization of RIS technology in wireless communication systems. Additionally, the study focuses on BPSK modulation, where the transmitted symbol can only have values of −1 and +1. This simplification overlooks the potential advantages or limitations of other modulation schemes that offer higher spectral efficiency or error resilience. Further investigations are required to extend the analysis for quadrature phase shift keying (QPSK) or higher-order quadrature amplitude modulation (QAM) modulations. Also, it is important to note that the proposed analytical framework relies on independent channel coefficients. However, extending the analytical framework to account for channel correlation is an open research direction that requires further investigation. Finally, extending the proposed framework to more realistic scenarios that involve multiple users with multiple antennas would be beneficial to further advance our understanding and practical utilization of RIS technology in wireless systems. By addressing these open research directions, we can continue to improve the performance and efficiency of RIS-aided wireless networks. Table 9 provides a comprehensive summarization of the studies discussed throughout this section.
2) RIS-Assisted Wireless Networks of Multi-Antenna Transmitter
In this section, we provide an in-depth review of recent studies in performance evaluation and parametric studies of RIS-aided networks with multi-antenna transmitters. The studies covered in this section are summarized in Table 10.
In [66], the authors have considered a RIS-assisted downlink multi-user MISO communication system. The system under investigation consists of one BS equipped with multiple antennas that serve multiple users via an RIS. Each user is equipped with a single antenna, whereas the RIS can be either one of these three types; reflective, transmissive, or hybrid. In the reflective type, the users are located on the same side of the BS, where the signals transmitted from BS are reflected to the users via RIS. In the transmissive type, the users are located on the opposite side of the BS and signals can pass through the RIS to serve users while in the hybrid type, the RIS has a dual function of the other two types. As shown in Figure 8, the users that are located on the reflection zone (i.e., on the same side of the BS) are called reflective users whereas, the users that are located on the transmission zone (i.e., on the opposite side of the BS) are called transmissive users. The authors have assumed that the direct transmission links between BS and the users are blocked while the RIS channels are modeled by considering path loss, fast fading, and RIS responses. For the considered system, the authors have derived an upper-bound approximation of the system sum rate based on CLT and Jensen’s inequality. Consequently, a discussion about which type of RIS will provide the best performance is reported. Monte Carlo simulation has been carried out and validates the theoretical analysis. The findings suggest that the choice of RIS type should be based on the specific system requirements and deployment scenarios. For example, when both the users and the BS are far from the RIS, the reflective or transmissive type of RIS is preferable. On the other hand, the hybrid type showed advantages when the users and the BS were both close to the RIS. These implications can guide network designers in selecting the most suitable RIS type based on the system characteristics. Besides, the study demonstrated that for a given number of RUs and irrespective of the RIS-users and BS-RIS distances, as the number of transmissive users increases, the system sum rate for the transmissive RIS type becomes larger, while the reflective RIS type’s capacity decreases. This insight can be valuable for optimizing system performance by strategically configuring RISs in multi-user MISO communication systems. Furthermore, the study suggests that, for a sufficiently large number of RUs, the hybrid RIS type consistently outperformed the other two RIS types for any arbitrary number of transmissive users. This implies that as the number of RUs increases, the hybrid type becomes a favorable choice for achieving higher sum rate performance. This highlights the scalability and potential of RIS technology in improving network capacity and accommodating larger user populations.
![FIGURE 8. - Illustration of the system model considered in [66].](/mediastore/IEEE/content/media/6287639/10005208/10356072/mahmu8-3342320-large.gif)
The implications of these results are crucial in guiding RIS type selection, optimizing system performance, highlighting RIS scalability, and providing practical insights for deploying RIS in wireless communication networks. However, these results are based on a simplified model for the RIS. For instance, the study assumes a simplified channel model by considering path loss, fast fading, and RIS responses. However, the real-world wireless channel is much more complex and dynamic, involving various factors such as multipath fading, interference, and shadowing. Ignoring these complexities may limit the accuracy of the results and their applicability to real-world scenarios. Moreover, the study assumed that the user has only one single antenna. In practical scenarios, users may have multiple antennas, enabling techniques like spatial diversity and beamforming. By assuming single-antenna users, the study overlooked the potential benefits and challenges associated with multi-antenna users. Furthermore, the study assumed a uniform amplitude response for the RUs ignoring the potential impact of amplitude variations on system performance. In reality, RUs may not have identical characteristics, and their amplitude responses could vary due to fabrication imperfections, positioning variations, or other factors. Thus, future studies could consider non-uniform amplitude responses among RUs and analyze the impact of amplitude variations on the overall system performance. In addition, the study assumed idealized RIS operation, where no power is dissipated by the RUs, and the signal energy is equally split between reflection and transmission in the hybrid type. These assumptions may not hold in practical RIS implementations, where power dissipation, losses, and inefficiencies can significantly affect system performance. Moreover, the study assumed that the distances between the reflective users and RIS are the same as the distances between the transmissive users and the RIS. In practical scenarios, these distances can vary due to their locations, mobility, or environmental factors. By assuming equal distances, the study overlooked the potential impact of varying distances on system performance. Hence, analyzing the optimal RIS type for different user-RIS distances would provide a more realistic and comprehensive evaluation of the system. Also, it should be noted that the proposed theoretical framework is built upon the assumption of a large number of RUs, invoking the CLT. Consequently, the applicability of the proposed framework is limited, as the obtained results may not be suitable for scenarios with a small number of RUs. Therefore, future research should focus on validating the applicability of the achieved results across various RIS configurations, encompassing scenarios with varying numbers of RUs. Finally, the study assumed a specific RIS plane positioning configuration where the RIS is positioned vertically to the ground, with one edge parallel to the ground and aligned with the direction from the BS to the RIS. However, this assumption overlooked the possibility of different orientations and angles of deployment. In reality, RISs can be positioned at various angles and orientations to achieve optimal signal reflection and performance. Therefore, further research should explore alternative RIS plane positioning strategies that take into account different angles and orientations for improved signal optimization and system performance.
In [67], the authors explored a RIS-aided multi-user downlink MISO communication system with ZF precoding. The system involved a single BS equipped with multiple antennas serving multiple single-antenna users via an RIS. The direct transmission links are assumed to be blocked, and the RIS channels are modeled considering path loss, fast fading, and RIS response. Assuming perfect CSI assumption at the BS, and using CLT and Jensen’s inequality, the authors derived an upper bound approximation of the system capacity. Monte Carlo simulation has been carried out and validates the theoretical analysis. The results showed that the system capacity is upper-bounded and cannot increase indefinitely as the number of RUs increases. In particular, for an infinite number of RUs, the system capacity can reach up to 71.5 bits/sec/Hz. Moreover, results demonstrated that using an excessively large number of RUs (
The study provides valuable insights for optimizing RIS design, achieving cost-effective deployment, and enhancing the overall performance of RIS-aided networks. Nevertheless, it is crucial to note that the proposed framework is based on CLT which gives accurate approximation only for a large number of RUs. Thus, the applicability of the achieved results is limited. This highlights the need for further investigation into alternative methods that can provide accurate approximations for smaller numbers of RUs. Moreover, the achieved results are based on ideal system model assumptions, such as perfect CSI, continuous phase shifts at the RIS, and fixed power allocation at the BS. While these assumptions allow for tractable analysis, they may not fully capture real-world scenarios. To further improve the understanding of RIS-aided networks, future research should investigate the impact of imperfect CSI, limited phase resolutions at the RIS, and dynamic power allocation at the BS. By doing so, we can gain a more accurate understanding of the system performance in real-world scenarios. Furthermore, the channel model used in the study may not capture the full complexity of real-world wireless environments, which can involve various factors such as multipath fading, interference, and shadowing. Ignoring these complexities may limit the accuracy of the results and their applicability to real-world scenarios. Additionally, the study highlights the importance of optimizing the trade-off between the number of RUs and the number of antennas at the BS in achieving low-cost deployment. Therefore, future research should address this problem and develop joint optimization frameworks that consider this trade-off and the achievable system performance. Lastly, although the study offers valuable insights into the effects of various design parameters, it fails to consider other parameters, such as the placement of RIS and the position of the RIS plane. Thus, further research is necessary in this area.
In [68], the authors explored the use of RIS in a downlink MISO communication system. The system involved a multi-antenna BS serving a single-antenna user with the assistance of an RIS. The direct transmission link was assumed to be blocked, leaving only the BS-RIS-user link available for communication. The authors assumed a flat fading scenario, where the BS-RIS channel consisted solely of a LoS component, while the RIS-user channel experienced Rayleigh fading. Additionally, the study assumed that the effective BS-RIS-user channel was perfectly known at the BS. To this end, the authors derived a closed-form expression for the maximum received SNR using the optimal active and passive beamforming at the BS and the RIS, respectively. The statistical properties of the received SNR were then characterized using both CLT and gamma approximations. Utilizing these statistical properties, the authors evaluated the system performance by deriving closed-form approximations for key metrics such as OP, average achievable rate, and ASEP. The two distribution approximations were assessed by comparing them to the exact distribution obtained through Monte Carlo simulations. The results showed that both approximations are accurate for a large number of RUs, with the Gamma distribution performing better in scenarios with a small number of RUs. As for the impact of different design parameters on system performance, findings revealed that increasing the number of RUs had a more significant impact on system performance compared to increasing the number of BS antennas. For instance, when the number of RUs was increased from 16 to 36 while keeping the number of BS antennas fixed at 16, the OP performance improved by approximately 9 dB, while the ASEP improved by 8 dB. On the other hand, increasing the number of BS antennas from 16 to 36 while maintaining a fixed number of 16 RUs resulted in an OP performance improvement of around 4 dB, and the ASEP improved by 3 dB. This highlights the greater influence of increasing the number of RUs on enhancing system performance compared to increasing the number of BS antennas.
The study offers valuable insights into the benefits and performance of RIS-assisted MISO communication systems. However, it is crucial to acknowledge several limitations. Firstly, the study is based on a simplified channel model that neglects the potential contributions of other channels and propagation effects, such as multipath fading, shadowing, and interference, which are commonly present in wireless communication systems. The simplified channel model may not accurately represent real-world scenarios, limiting the generalizability of the study findings. Additionally, the study derives optimal beamforming solutions based on the assumptions made. However, it is important to note that the optimality of these solutions is highly dependent on the validity of the assumptions. Any deviations from the assumed conditions, such as variations in channel models or channel knowledge, may result in the derived optimal solutions becoming suboptimal or even infeasible. For example, if the channel between the BS and RIS experiences fading instead of LoS, it would impact the optimal beamformer and system performance. Thus, addressing this issue and determining the optimal beamformer and system performance under fading conditions for both the BS-RIS and RIS-user links are important areas for future research. Moreover, practical implementation considerations such as imperfect CSI, hardware constraints, and power consumption are not addressed. The study focused on limited design parameters, neglecting factors like RIS placement and plane position. Furthermore, the findings may not be generalized to different scenarios or multi-antenna users. Considering these limitations is crucial when interpreting the achieved results, and further research is needed to address these limitations and explore the practical implications of RIS-assisted MISO communication systems.
B. Multiple RISs-Assisted Wireless Systems
In this section, we delve into the performance analysis and parametric studies of wireless networks enhanced by the utilization of multiple RISs. To enhance comprehension of this section, we have provided a summary of these studies in Table 11.
The authors in [69] have considered a point-to-point downlink SISO with a RIS-aided system over Nakagami-
![FIGURE 9. - Illustration of the RISs distributions considered in [69].](/mediastore/IEEE/content/media/6287639/10005208/10356072/mahmu9-3342320-large.gif)
Monte Carlo simulation has been conducted and emphasizes the correctness of the proposed framework. Results demonstrated that increasing either the number of RUs or the number of RISs results in a lower OP and improved performance. Both parameters have almost equal positive effects on error performance. The authors also analyzed the joint impact of the number of RUs, number of RISs, and distances between the source-RISs and RISs-destination on performance. To do this, they considered different network configurations where the number of RUs, the number of RISs, and their location and distribution were varied among different scenarios. The results showed that assigning a fixed number of RUs to a single RIS leads to lower outages than distributing them equally among multiple RISs. Additionally, the optimal location for the RIS was found to be at the midpoint between the source and destination. For the multi-RIS setup, serial RIS configuration was found to yield better outage performance than parallel configuration. Furthermore, unequal distribution of RUs among RISs yielded better error performance compared to equal distribution. Lastly, the study found that increasing the channel shape parameter m led to better OP and ASEP performance.
This study provides valuable insights into the design considerations for RIS-aided systems. By analyzing various network configurations and placement strategies, such as assigning a fixed number of RUs to a single RIS, placing RIS at optimal distances, and considering serial RIS configurations and unequal distribution of the RUs among RISs, the study offers guidance for optimizing the performance of RIS-aided systems in real-world deployment scenarios. However, it is important to note that the study assumes a simplified system model that may not fully capture the complexities and diversity of real-world scenarios. Real-world systems often involve multiple antennas, multi-user scenarios, and interference from neighboring cells, which can impact the performance of RIS-aided systems differently. Therefore, extending the study to multi-user MIMO scenarios where multiple sources and destinations communicate through RISs and investigating the impact of the number of BS antennas, number of RIS RUs, and number of RISs and their trade-offs would be beneficial. Moreover, the selection of a single RIS from a set of available options based on the highest SNR is built upon pilot estimation that assumes reciprocal uplink/downlink channels, meaning that the channel characteristics are the same in both directions. However, in practical scenarios, uplink and downlink channels can exhibit asymmetry due to factors such as antenna orientation, scattering environment, and interference. Ignoring this asymmetry may lead to inaccurate performance evaluations. Furthermore, it would be valuable to enhance the investigation by considering the impact of channel estimation error and phase noise. By extending the study to incorporate these practical limitations, a more comprehensive understanding of the performance degradation caused by these limitations can be achieved. This knowledge is crucial for developing effective solutions to mitigate their effects. Additionally, the path-loss model used in the study assumes that the overall path loss of the RIS-assisted path is determined solely by the distances between the transmitter, RIS, and receiver, and can be represented as the product of independent path losses in the first and second hops. However, in reality, the path losses are influenced by various factors and dependencies that are not accounted for in this simplified model. These factors include shadowing, multipath fading, and other environmental effects that can significantly impact signal propagation. Moreover, the study assumes ideal RISs with full reflection capability, with an efficiency of one implying that all the power incident on the RIS is perfectly reflected. This assumption overlooks imperfections and losses that are present in real-world RISs and can affect efficiency. Neglecting these factors can lead to an overestimation of system performance. Therefore, a more comprehensive and realistic model should consider these additional factors and accurately characterize the path loss and RIS efficiency. By addressing these limitations, a deeper understanding of the system’s behavior can be gained, leading to more effective design and optimization strategies for multi-RIS deployments in practical settings.
The authors in [186] have evaluated the performance of a multi-RIS-aided downlink SISO wireless network in the presence of direct links and with discrete phase shifts at the RISs. The system involves a single-antenna source that serves a single-antenna destination with the assistance of L RISs. The RISs are distributed uniformly in a serial manner between the source and the destination. That is, the RISs are placed along a hypothetical straight line connecting the source and destination with equal separating distances between each other, as depicted in Figure 10. Each RIS consists of an equal number of passive RUs N
![FIGURE 10. - Illustration of the system model considered in [186].](/mediastore/IEEE/content/media/6287639/10005208/10356072/mahmu10-3342320-large.gif)
The findings indicate that for a fixed number of RUs, the outage performance improves as the number of RISs increases. For instance, when using a single RIS with 32 RUs, an average transmit SNR of 1 dB is required to achieve an OP of
The study contributes to the understanding of the performance characteristics of multi-RIS-aided wireless networks while taking into account limited phase resolution. It also highlights the advantages of utilizing distributed RISs and identifies crucial factors that affect system performance and optimization. Nevertheless, it is important to note that the study relies on simplified network configurations to facilitate mathematical modeling and analysis. These configurations may not fully capture the complexities and variations present in real-world wireless networks. Therefore, the findings and conclusions may not be directly applicable to all possible deployment scenarios. Thus, it is necessary to extend the analysis to more realistic network topologies such as multi-user MIMO scenarios. The design and optimization of multi-RIS in such scenarios can be challenging and require further investigations. Additionally, although the study has considered some practical limitations, such as discrete phase shifts at the RIS and special correlation between channels, other sources of imperfections that can affect the performance of distributed RISs have not been taken into account. Imperfect CSI, hardware impairments such as non-linearities, and power constraints are important factors that should be investigated in future work. In addition, the study assumes the placement of the source and destination in the far field of the RISs. This assumption simplifies the analysis but may not hold in practical scenarios. In reality, the source and destination may be located at varying distances from the RISs, leading to different propagation conditions and signal characteristics. Moreover, the study assumes that the RISs are uniformly distributed serially between the source and destination with equal RUs. However, this assumption may not accurately reflect practical deployment scenarios, where the placement of RISs could be influenced by various factors such as physical obstructions and resource availability. As a result, it is crucial to investigate the impact of non-uniform RIS distribution on system performance and to explore optimized RIS placement strategies. Further research is necessary to address these important considerations. Furthermore, the primary focus of this study is on distributed passive RISs. However, there is potential for additional performance gains through the use of hybrid active-passive RISs. These systems incorporate active RUs that can transmit signals. As such, extending the design and analysis of hybrid multi-RISs is an open research direction that requires further investigation. Finally, it is important to note that the statistical characterization of the proposed theoretical framework relies on the CLT approximation. Consequently, the applicability of the achieved results is limited to scenarios with a large number of RUs. Therefore, future research should focus on developing more accurate analytical models to evaluate the performance of the considered system, particularly in scenarios with a small number of RUs.
The authors of [187] have explored a system model similar to that of [186], in which a single-antenna source communicates with a single-antenna destination with the aid of multiple RISs over a Nakagami-
The implications of the study and its achieved results highlight the potential of multi-RIS systems in improving wireless communication performance, energy efficiency, and system reliability. These findings provide insights into the design, optimization, and utilization of RISs in practical wireless networks. Nevertheless, it is important to acknowledge the limitations of the study. One of the main limitations is the use of a simplified model that considers SISO network configurations with a single antenna at the transceivers. While this model provides valuable insights, it may not fully capture the complexities of real-world scenarios, which often involve multiple users with multiple-antenna transceivers. Therefore, expanding the study to incorporate multi-user MIMO would provide more accurate insights into the system performance and enhance our understanding of real-world implementation. Moreover, the study assumes i.i.d. Nakagami-
Future Research Prospects for Performance Analysis and Parametric Studies
In the preceding section, we have outlined various research gaps related to each study. However, our attention will now be directed toward broader research directions that emerge from the collective insights of the analyzed studies in Section VI. Figure 11 offers a visual illustration of the future work and prospects discussed in this section.
A. Realistic Deployment Scenarios
As previously discussed, the majority of research on RIS-assisted systems has focused on SISO setups with single-user deployment [64], [165], [166], [171], [172], [175], [178], [181], [182]. This approach allows for simplified analysis and mathematical modeling, providing a baseline understanding of the capabilities and limitations of RIS technology. By isolating the impact of the RIS on a single user, researchers can investigate the effects of key parameters such as the number of RUs, RIS placement and plane positioning, and phase-shifting capabilities on system performance. Additionally, studying SISO deployments helps to characterize the gains achieved solely through the RIS without the interference of multi-user interactions or multiple antennas. This enables a more detailed examination of the benefits provided by the RIS, such as enhanced signal strength, improved coverage, or reduced path loss. Understanding these advantages is crucial for assessing the potential of RIS technology and optimizing its deployment. However, it is important to note that evaluating performance solely in a SISO scenario may not capture the full potential of RIS-aided networks. Real-world wireless communication systems often involve multiple users sharing the same resources and multiple antennas at the transmitters and receivers. Ignoring these factors can lead to a limited understanding of the actual system behavior and performance in practical deployment scenarios. Therefore, it is crucial to extend the analysis to multi-user and multi-antenna scenarios. This will provide a more comprehensive evaluation of RIS-aided networks, accounting for the impact of interference, multi-user interactions, and the potential gains offered by multiple antennas and coordinated beamforming techniques. While some studies have considered multi-user deployment in SISO and MISO configurations [65], [66], [67], MIMO deployments have received less attention. Therefore, it is essential to expand the analysis to include multi-user deployment in MIMO scenarios. Furthermore, the impact of mobility, including user mobility and RIS mobility, must be taken into account to assess the system’s performance under realistic dynamic scenarios. This will provide a more complete understanding of the potential benefits of RIS technology and enable the optimization of its deployment in practical wireless communication systems. Additionally, far-field transmission scenarios have been a key focus in the current studies. Although these scenarios are relevant for long-range communication, they may not adequately address the unique requirements and challenges posed by short-range communication in near-field scenarios. Therefore, it is imperative to explore the applicability of RIS technology in near-field transmission scenarios, as it holds significant potential for various wireless communication applications.
B. Practical Considerations
While existing studies have provided valuable insights into the theoretical aspects of RIS technology, they have largely relied on ideal assumptions that do not reflect the practical challenges of deploying this technology. For instance, a wide range of previous studies has assumed perfect knowledge of channel characteristics, unlimited phase resolution at the RIS, and the absence of hardware imperfections. However, in reality, these assumptions are not always valid, and further research is needed to address the practical deployment challenges of RIS technology. To this end, future research should focus on investigating the feasibility and scalability of RIS implementation, considering cost-effective RIS design, and evaluating the impact of hardware imperfections and other practical constraints such as imperfect CSI estimation and limited phase resolution with noise at the RIS. Additionally, most current studies assume that all RIS unit elements are identical and have equal power levels, which is not always the case in practical scenarios. Therefore, future research should explore how variations in element characteristics, such as different gain or phase response, and power levels affect system performance and develop techniques to mitigate their impact. One potential solution to these challenges is to develop robust, adaptive power allocation algorithms or beamforming techniques that can compensate for element variations. By addressing these practical considerations, future research can pave the way for the successful integration and deployment of RIS-aided systems in real-world scenarios. Ultimately, this will enable us to fully realize the potential of RIS technology and unlock new opportunities for wireless communication. Lastly, the majority of the studies assume ideal RIS with full reflection capability and efficiency, meaning that the power incident by RIS is perfectly reflected, ignoring the losses and imperfections that RIS might produce in the reflecting signal. However, in practice, RIS RUs may have non-ideal reflection properties due to manufacturing imperfections or environmental factors. Therefore, considering these imperfections and losses in the RIS models would provide a more comprehensive understanding of system performance for practical implementation.
C. Channel State Estimation
Optimizing the parameters of RIS is important for achieving optimal system performance. The channel response plays a significant role in this optimization process, as it encompasses various factors such as fading, scattering, and shadowing. However, obtaining accurate CSI is essential in RIS-aided wireless communications and can be challenging in practice [42]. The flexibility of the served clients and the presence of obstacles can hinder the continuous attainment of accurate CSI values, which poses a problem for optimizing network performance when dealing with unreliable CSI. To enable real-time and effective RIS-assisted transmission, it is essential to address the issue of accurately estimating the CSI value and optimizing the network, particularly when dealing with poor CSI conditions. One potential solution is the use of RL approaches [11], which have shown promising results in mitigating the impact of poor CSI. However, RL approaches can require extensive training time due to the constant changes in channel conditions. Therefore, further research is needed to explore accurate and robust channel estimation techniques that minimize training overhead and mitigate the effects of estimation errors.
D. RIS Placement and Plane Positioning Design
The placement of RIS is of great importance in addressing signal transmission challenges in obstructed areas and minimizing distortion and interference in signal propagation [64], [186]. The effectiveness of signal reflection and the achievement of optimal transmission rates are directly impacted by the placement of an RIS [169], [189]. Therefore, careful consideration of RIS placement is necessary to achieve a high optimal achievable rate. Furthermore, the physical size of the RIS and the number of RUs employed also play a significant role in system performance [166]. Increasing the size of the RIS and the number of RUs can enhance the percentage of signal reflection in the desired direction, but it may also incur higher overhead costs [190], [191]. To optimize the performance of RIS, it is essential to strike a balance between the size of the RIS, the number of RUs, and the placement. By doing so, we can achieve the highest possible transmission rates while minimizing costs. Moreover, the authors in [166] emphasize the significant role of positioning the RIS plane. However, the existing literature lacks adequate attention to this crucial design parameter. Hence, future research should focus on exploring the joint impact of RIS placement, RIS plane position, and other relevant parameters such as the RIS size and number of RUs. By investigating these factors comprehensively, it will be possible to optimize system performance effectively.
E. Multiple RIS Deployment
Most studies conducted so far have primarily focused on analyzing the performance of a single RIS. Nevertheless, there exist a limited number of works in the literature that have considered the case of multiple RISs [69], [186], [187], [188], [192], [193], [194], [195]. While these studies have provided valuable insights into system design and parameters, they have often relied on simplified system models that incorporate SISO scenarios, which may not fully capture real-world complexities. Incorporating the concept of multiple RISs in multi-user MIMO systems and studying the impact of various system parameters would be a valuable extension. This research direction would involve investigating the effects of factors such as the number of RISs, the number of RUs within each RIS, the distribution of these elements across the RISs, the positioning of the RISs in the system plane, as well as the number of BS antennas and the associated trade-offs. Exploring the impact of different system parameters and their trade-offs in this context would provide valuable insights for the design and optimization of future wireless communication systems.
F. Quantization Schemes
In the studies discussed so far, which considered discrete phase shift resolution at the RIS, a common assumption has been the use of uniform quantization for channel phase shifts with a fixed number of levels. However, in practical scenarios, the number of quantization levels may vary due to implementation and hardware constraints. To address this, future research needs to explore adaptive quantization schemes that can dynamically adjust the quantization levels based on changing channel conditions. This research direction would involve investigating the impact of adaptive quantization on system performance and comparing them to fixed-level quantization. By doing so, we can develop techniques that adapt to the changing channel conditions, thereby improving system performance and robustness. This adaptive approach is particularly important in dynamic wireless environments where channel conditions may fluctuate due to factors such as mobility, interference, and varying user densities.
G. Performance Metrics
Previous studies have primarily focused on analyzing the performance of RIS-based systems using common performance measures like OP, BER, achievable rate, and ASEP. These metrics have provided valuable insights into the overall system performance, allowing for the effective evaluation of various system designs and techniques. However, one important performance metric that has been largely overlooked in the literature is latency. In future research, it is crucial to include latency as a performance metric, as it plays a critical role in many applications, particularly those requiring real-time communication and responsiveness. Considering latency as a performance metric would provide a more comprehensive understanding of the system’s capabilities and limitations, enabling the development of optimized designs and techniques that meet the demanding latency requirements of different applications. Moreover, considering multiple metrics and combining them into an integrated metric can offer a more informative and comprehensive evaluation of RIS-aided systems. This integrated metric serves as a valuable tool in understanding the trade-offs between different performance aspects and provides a comprehensive overview of the system’s capabilities and limitations. However, creating such an integrated metric requires careful consideration of the relative importance of individual metrics, which should be tailored to the specific application requirements. In other words, the selection and combination of metrics should be application-specific, as different applications may demand varying aspects. For instance, in real-time communication systems, online gaming, or virtual reality, latency may emerge as the most critical metric. On the other hand, for IoT networks, a balanced combination of power efficiency, error rates, and latency may hold greater significance. In future research, developing an integrated metric that considers relevant performance metrics will certainly contribute to a more comprehensive understanding of RIS-based systems. This will allow for the design and optimization of RIS-based communication systems that meet the diverse and demanding needs of real-world applications.
H. Channel Modeling
The majority of the studies on RIS-based systems have primarily utilized simplified fading models to represent the channel, neglecting the complexities of real-world wireless channels affected by path loss, shadowing, and multipath fading. To gain a comprehensive understanding of RIS performance in realistic environments, future research should incorporate more comprehensive channel models that consider these factors. Additionally, the impact of other fading models, such as frequency-selective fading, time-selective fading, and fast fading, should be explored. Moreover, the majority of studies have assumed i.i.d channels, disregarding the presence of channel correlation. However, in practical wireless communication scenarios, channels often exhibit correlation due to factors such as spatial proximity, scattering, and antenna configurations. Ignoring channel correlation can lead to an oversimplified representation of the system behavior and may not reflect real-world performance accurately. Hence, future research should take into account channel correlation and investigate its impact on the performance of RIS-assisted systems to provide a more realistic analysis.
I. Analytical Approaches
The majority of studies on RIS-based systems have primarily relied on CLT-based approximation [65], [66], [67], [172], [175], [178], [186], while others have explored the use of the Gamma approximation [68], [69], [166], [187]. It is important to note that while the CLT approximation guarantees accurate results when there is a large number of RUs, its reliability diminishes when the number of RUs is small [170]. Furthermore, the CLT approximation may exhibit poor performance in high SNR scenarios [174]. Consequently, the obtained results may not apply to scenarios with a small number of RUs or high SNR. In contrast, the Gamma-based approximation offers wider applicability. Unlike the CLT approximation, the accuracy of the Gamma-based framework is not constrained by the number of RUs, providing a more general representation [170]. However, it should be noted that the Gamma-based framework may not be suitable for diversity analysis, as it cannot fully extract the diversity order even in the absence of phase errors [170]. To address these limitations, future research should prioritize the investigation of more sophisticated and precise analytical models that can accurately assess network performance under high SNR conditions and for scenarios with a small number of RUs. These analytical models should be designed to accurately capture the diversity order performance of RIS-based systems. By addressing these research gaps and exploring alternative modeling approaches, we can enhance the accuracy and applicability of performance analysis in RIS-based systems.
Conclusion and Future Work
A. Conclusion
RIS is a promising technology that enhances signal strength and energy efficiency. It enables path customization, and interference mitigation, making it a cost-effective solution for high-interference environments. However, efficient optimization and performance evaluation are important in fully unlocking its full potential and understanding optimal configurations for real-world deployments. This work uniquely integrates recent studies on RIS beamforming optimization designs and performance evaluation, providing a comprehensive and insightful assessment that sets it apart. The review encompasses state-of-the-art techniques and methodologies used in these areas, highlighting key advancements in the field. It emphasized practical limitations and challenges associated with implementing RIS technology in real-world settings. This practical perspective is crucial for bridging the gap between theoretical advancements and practical applications. As a result, this work serves as a valuable resource for both academics and practitioners seeking a complete understanding of the technology. In particular, we offered a detailed and comprehensive discussion of the limitations and potential solutions of each of the reviewed works in joint beamforming optimization and performance evaluation of RIS-aided systems. Such discussion guides to practitioners and researchers on implementing RIS in real-world. We identified research gaps and challenges, emphasizing the need for future investigations in key areas. These areas include considering practical limitations and real-world deployment scenarios. Moreover, we highlighted the importance of developing robust algorithms that can handle real-world imperfections and exploring optimization techniques that can guarantee global optima solutions with computational efficiency. Furthermore, we emphasized the significance of investigating analytical approaches for accurate network performance assessment under diverse conditions, advancing multi-objective joint optimization, examining the joint impact of relevant RIS design parameters, and developing application-specific integrated performance metrics. Precise channel modeling, accurate channel state estimations, and adaptive quantization schemes for limited phase resolutions are also important aspects that require further investigation. The findings of this research hold potential for various practical applications. They can be directly applied in scenarios where RIS technology aims to enhance wireless communication, such as urban environments with high signal interference, indoor spaces with poor coverage, and smart transportation systems that require reliable and efficient connectivity. Additionally, the insights gained from this research can guide the deployment of RIS in emerging technologies like IoT, autonomous vehicles, and smart cities. The direct application of the findings of this review in the near term would be to the field trials inside the operational 5G wireless networks to realize and maximize the proposed gains from adding RIS-related approaches, methods, and algorithms compiled by this review.
B. Future Work
Expanding the research to consider other factors or parameters not covered in the current study can broaden its scope and applicability. Researchers can explore several avenues for extension. One approach could be investigating a wider range of performance metrics that are relevant to specific real-world applications. For instance, metrics related to latency, or QoS, and the newly introduced quality of experience (QoE) could be incorporated to provide a more comprehensive evaluation, aligning the findings of the study with the practical requirements of diverse real-world applications. Moreover, the parameters of the algorithms for modeling and optimization can also be covered to study their effects on their level of accuracy and computational complexity. This comprehensive approach not only enhances the research relevance but also empowers stakeholders with valuable insights, aiding them in making informed decisions for the implementation of RIS-aided systems. Furthermore, exploring the security aspects of RIS-aided networks, assessing vulnerabilities, and proposing strategies for secure real-world deployment is also a valuable extension. Additionally, several emerging technologies and advancements hold the potential to enhance the performance of RIS-based systems. Advancements in metamaterials and nanotechnology can lead to the development of more efficient RIS elements, enabling accurate control over the reflected signals and improved beamforming capabilities. Moreover, the integration of AI and ML techniques can enable RIS-aided systems to adapt and optimize their configurations in real time based on changing network conditions and user requirements, leading to increased efficiency and performance.
Furthermore, advancements in mmWave and THz communication technologies can expand the frequency ranges in which RIS can operate, offering new possibilities for signal manipulation and propagation enhancement. Finally, the evolution of 6G and beyond wireless communication standards will likely incorporate RIS technology as a fundamental component, driving research and development efforts to make RIS more compatible and efficient in future networks. These emerging technologies collectively promise to further unlock the potential of RIS-based systems and enable innovative applications across various domains.