Key points are not available for this paper at this time.
In this paper, we concentrate on a reconfigurable intelligent surface (RIS)-aided mobile edge computing (MEC) system to improve the offload efficiency with moving user equipments (UEs). We aim to minimize the energy consumption of all UEs by jointly optimizing the discrete phase shift of RIS, UEs' transmitting power, computing resources allocation, and the UEs' task offloading strategies for local computing and offloading. The formulated problem is a sequential decision making across multiple coherent time slots. Furthermore, the mobility of UEs brings uncertainties into the decision-making process. To cope with this challenging problem, the deep reinforcement learning-based Soft Actor-Critic (SAC) algorithm is first proposed to effectively optimize the discrete phase of RIS and the UEs' task offloading strategies. Then, the transmitting power and computing resource allocation can be determined based on the action. Numerical results demonstrate that the proposed algorithm can be trained more stably and perform approximately 14% lower than the deep deterministic policy gradient benchmark in terms of energy consumption.
Li et al. (Tue,) studied this question.