Exploiting Semantic Communication for Non-Orthogonal Multiple Access

A novel semantics-empowered two-user uplink non-orthogonal multiple access (NOMA) framework is proposed for resource efficiency enhancement. More particularly, a secondary far user (F-user) employs the semantic communication (SemCom) while a primary near user (N-user) employs the conventional bit-based communication (BitCom). The fundamental performance limit, namely semantic-versus-bit (SvB) rate region, of the proposed semantics-empowered NOMA framework is characterized. The equivalent SvB rate region achieved by the conventional BitCom-based NOMA is provided as the baseline scheme. It unveils that, compared to BitCom, SemCom can significantly improve the F-user’s performance when its permitted transmit power is strictly capped, but may perform worse when its permitted transmit power is high. Guided by this result, the proposed semantics-empowered NOMA framework is investigated over fading channels. An opportunistic SemCom and BitCom scheme is proposed, which enables the secondary F-user to participate in NOMA via the most suitable communication method at each fading state, thus striking a good tradeoff between its own achieved performance and the interference imposed on the primary N-user. Two scenarios are considered for employing the opportunistic scheme, namely on-off resource management and continuous resource management. For each scenario, the optimal communication policy over fading channels is derived for maximizing the ergodic semantic rate achieved at the secondary F-user, subject to the minimum ergodic bit rate constraint of the primary N-user. Numerical results show that: 1) proposed opportunistic scheme in both scenarios can achieve higher communication performance for NOMA than the baseline schemes merely employing SemCom or BitCom; 2) SemCom can better guarantee the performance of the F-user admitted in NOMA than BitCom when the communication requirement of the primary N-user is high; and 3) continuous power control at the F-user is necessary for ensuring high performance over fading channels, while the on-off time scheduling is sufficient.

I. Introduction

Thanks to the rapid development of wireless communications, people’s life and work styles have been significantly improved. Compared to the earliest wireless communication technology in the 1920s, the current 5G wireless communication system enables people to communicate with each other through speech, text, and video in a ubiquitous manner. Reviewing the historical development of wireless communications, the Shannon classical information theory [3] provides the fundamental rate limit for realizing reliable communications. In the Shannon paradigm, the design metric is to ensure the bit sequences converted from the source can be accurately transmitted to the destination, where the intrinsic meaning of the source is put aside during the transmission. Guided by the Shannon paradigm, “engineering” problems can be formulated considering generalized performance metrics, such as symbol-error rate (SER) and bit-error -rate (BER), which thus receive growing research efforts in the past a few decades. Despite a series of sophisticated communication technologies [4], [5], [6] having been proposed for approaching or achieving the Shannon rate limit, the communication requirements have not been completely fulfilled, which calls for the development of new communication technologies.