STAR-RIS-Aided Full-Duplex ISAC Systems: A Novel Meta Reinforcement Learning Approach

In this work, we consider a full-duplex (FD) communication system that uses a simultaneous transmission and reflection (STAR) enabled reconfigurable intelligent surfaces (RIS) to assist the communication and sensing between a base station (BS) to a single set of UL and DL user, and target over the same-time frequency dimension. In order to explore the performance of the proposed framework, we offer an analytical framework and accordingly, we propose an optimization problem to jointly optimize the phase-shift matrices at the STAR RIS (S-RIS) that maximizes the possible sum-rate. Due to the non-convexity of the optimization problem, we then propose a low-complexity meta-reinforcement learning (MRL) algorithm that reduces the overall training overhead. We also demonstrate the effectiveness of the proposed algorithm in providing near-optimal design in the case of imperfect channel state information (ICSI). Additionally, in order to verify how well the proposed framework work and to show the superiority of the proposed algorithm, we provide a fair comparison with two baseline schemes a) twin delayed deep deterministic policy gradient (TD3) and b) deep deterministic policy gradient (DDPG). Simulation results verify that the proposed approach results in superior performance.