Optimizing SWIPT in Multi-RIS Aided V2I Networks: A Deep Learning Approach

This paper investigates the effectiveness of employing multiple reconfigurable intelligent surfaces (RIS) for simultaneous wireless information and power transfer (SWIPT) in a vehicle-to-infrastructure (V2I) system. The optimal RIS is selected for transmission based on instantaneous signal-to-noise ratio (SNR) values, with the objective of optimizing the SWIPT system employing the power-splitting (PS) protocol and nonlinear energy harvesting (NL-EH). A unified objective is proposed to maximize information rate and harvested energy via joint optimization of transmit power and power splitting factor. Nonconvexity is addressed via an iterative algorithm, supported by closed-form expressions obtained through Karush-Kuhn-Tucker (KKT) conditions. Monte-Carlo simulations are performed to validate the accuracy of the analytical expressions. Additionally, a deep neural network (DNN) framework is introduced for realtime optimization prediction, achieving superior SWIPT performance over single RIS configurations with reduced complexity and faster execution.