Null-Space Based Design With Learning for RIS-Aided Wireless Information and Power Transfer

The paper presents a novel concept for reconfigurable intelligent surfaces (RIS) aided simultaneous wireless information and power transfer (SWIPT). The concept leverages, for the first time, the degrees of freedom provided by the null-space of the channel between the access point and the information receivers (IRs) through the RIS to send an additional energy signal to the energy receivers (ERs), simultaneously with the data transmission to the IRs. As with any new concept, the first step must be to validate it and assess its advantages and potential limitations in a typical wireless system composed of a multi-antenna transmitter, multiple IRs and ERs, and a RIS of varying dimensions. The paper does exactly this by designing a SWIPT system where the proposed null-space-based (NSB) approach is implemented. This design is then used to quantify the performance of the new approach and benchmark it against prior art. In this regard, we seek to maximize the ERs' harvested power and the IRs' data rate in the designed SWIPT system, which results in a multi-objective optimization problem proper to our NSB approach. To address this optimization problem, we propose and implement two solution approaches: one leveraging the deep deterministic policy gradient (DDPG) method, and the other utilizing alternating optimization (AO). We compare both solutions in terms of system performance, showing that the DDPG consistently outperforms the AO by 2–3% in all tested cases. Additionally, benchmarking against state-of-the-art approaches demonstrates that the proposed NSB SWIPT design outperforms existing SWIPT benchmarks by 15–20%, both with and without RIS.