High-Throughput Wireless Uplink Transmissions Using Self-Powered Hybrid RISs

This article investigates the uplink of a reconfigurable intelligent surface (RIS)-assisted wireless communication system. In this uplink, the RIS is powered by the energy harvested from ambient energy sources, and assists multiple users in uploading data to a multi-antenna base station (BS). A hybrid architecture is proposed for the RIS, so that each reflecting unit (RU) at the RIS is enabled to select its working mode among passive, active and deactivated modes. In this way, the RIS can schedule the energy in a fined-grained manner. Meanwhile, a new protocol is proposed to enable the RIS to schedule the harvested energy with a forward-looking approach. Under the hybrid RIS architecture and the newly proposed protocol, an optimisation problem is formulated to jointly optimise the working modes of RUs, the amplitude coefficient of active RUs, the receive beamforming at the BS, the power allocation at the users, with the goal to maximize the long-term system throughput by considering a minimum-rate-requirements constraint at each user and an energy scheduling constraint at the RIS. The formulated problem is an intractable dynamic and mixed-integer nonlinear programming. To solve this problem, a hierarchical deep reinforcement learning based framework is proposed. Simulation results show that, by using hybrid RISs, our self-powered wireless system can achieve up to 12 (5) times of the throughput than the throughput achieved by a self-powered wireless system with just active (passive) RISs and myopic energy scheduling.