Optimal Dynamic Resource Allocation for Multi-RIS Assisted Wireless Network: A Causal Reinforcement Learning Approach

This study explores optimal dynamic resource allocation in mobile reconfigurable intelligent surface (RIS)-aided wireless networks facing uncertain, time-varying channels. Unmanned aerial vehicles (UAVs) strategically deploy multiple RIS to enhance mobility. Addressing uncertainties, a novel causal reinforcement learning-based online dynamic resource allocation algorithm is proposed. It features a structural causal model for crucial parameters in time-varying channels and a Q-learning Adaptive Dynamic Programming algorithm optimizing mobile RIS deployment. A causal actor-critic algorithm further refines transmit power and phase shift control policies. Numerical simulations validate the approach's efficacy, underscoring its potential to maximize spectrum efficiency and enhance dynamic wireless network performance.