Task Offloading Strategy for Aerial RIS-Assisted Wireless-Powered MEC Based on Deep Reinforcement Learning

In this paper, we propose a task offloading strategy based on the aerial reconfigurable intelligent surface (RIS)-assisted wireless-powered mobile edge computing (WP-MEC) system. Specifically, by splitting RIS arrays within the geometric space, the aerial RIS simultaneously harvests energy from radio-frequency (RF) signals transmitted by the base station (BS) while also assisting users in offloading the portion task to the BS. Under this strategy, we study an user's energy consumption minimization problem of the system by jointly optimizing the time allocation, RIS's phase shifts, the RIS scheduling matrix, the users’ power, and the task offloading radio of users, while satisfying the constraints of the aerial RIS's energy consumption and the computational capability of users. Since the considerable optimization problem is highly non-convex, it is difficult to tackle directly. Hence, we propose a softmax deep double deterministic policy gradient (SD3)-based deep reinforcement learning (DRL) task offloading algorithm to tackle this problem. The simulation results confirm the effectiveness and superiority of the proposed algorithm.