Joint Optimization of Service Caching and Task Offloading for Customer Application in MEC: A Hybrid SAC Scheme

Mobile Edge Computing (MEC), with advantages in high bandwidth and low latency, enables the development of numerous promising commercial services on edge servers near users. However, complex associations among users, servers and services require nontrivial collaboration to boost the performance of heterogeneous applications with diverse service requirements. In this paper, we study a joint task offloading and service caching problem in commercial MEC networks, aiming to minimize the delay and computational cost for all users. To this end, we first formulate the above issue as a complex optimization problem, and decompose it into two sub-problems for reducing computational complexity while maintaining its accuracy. Then, we propose a data-driven Hybrid Soft Actor-Critic scheme, where the deep reinforcement learning-based part determines the near-optimal service caching decisions, and the convex optimization technology-based part calculates the optimal offloading decisions. Finally, simulation results show that our proposed scheme improves the performance of accuracy and convergence when dealing with high-dimensional action spaces, and it outperforms the baseline schemes.