Potential Game-Based Computation Offloading in Edge Computing With Heterogeneous Edge Servers

With the proliferation of mobile phones, IoT devices, and the rising demand for computational resources, computation offloading has emerged as a promising technique for improving performance, and optimizing resource usage. It involves transferring computational tasks from local devices to edge servers. However, reducing latency and device energy consumption remains a challenge in current research. In this paper, we propose a potential game-theoretic approach to optimize computation offloading in edge computing environments. We consider heterogeneous edge servers, where each server may have different computational capabilities. By formulating the problem as a potential game, we have end devices acting as players deciding whether to execute tasks locally or on edge servers. Our framework includes utility functions capturing the latency-energy consumption trade-off. Through a detailed analysis, we introduce an innovative algorithm for potential games aiming at achieving Nash equilibrium. This algorithm demonstrates exceptional convergence properties, ensuring reliable convergence even in complex scenarios. Extensive experiments validate the convergence of our algorithm and demonstrate its better performance compared to other benchmark algorithms in terms of latency and energy consumption.

I. Introduction

The exponential growth of mobile devices and the proliferation of IoT devices [1], i.e., devices used by end-users, have led to an unprecedented demand for computational resources [2]. These devices can offer a wide range of network services, such as data analysis, virtual reality, and video games. The explosive growth of network traffic has further intensified the demand for computational resources, placing greater pressure on the processing capabilities of devices. Although these devices are equipped with CPUs capable of processing data, the computational capabilities of these devices are still insufficient to meet the demands of computationally intensive tasks. Furthermore, these devices typically rely on battery power and need to efficiently manage their energy consumption [3], [4]. In this context, it becomes crucial to address the challenges posed by the increasing demand for computational resources while ensuring efficient energy utilization and minimizing latency, thereby enhancing application performance and user satisfaction to ensure a seamless and enjoyable experience with these devices and services.