Beamforming for STAR-RIS-Aided Integrated Sensing and Communication Using Meta DRL

We consider an integrated sensing and communication (ISAC) system, in which a simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) assists a base station in transmitting communication signals to mobile users and conducting sensing tasks toward specific targets. We formulate a transmit beamforming and phase shift optimization problem to jointly maximize the total communication data rate and total effective sensing power. The optimization problem is inherently non-convex, making it challenging to find an optimal solution. To tackle this difficulty, we propose a meta soft actor-critic (meta-SAC) algorithm, which is a fusion of the SAC algorithm and meta-learning techniques. Through extensive simulations, we demonstrate that the proposed meta-SAC algorithm outperforms traditional deep reinforcement learning methods, thus showing its potential to enhance the performance of ISAC systems significantly.