Joint Beamforming Design for Wireless Powered IRS Assisted Multi-User ISAC Systems

Although resorting intelligent reflecting surface (IRS) into integrated sensing and communications (ISAC) systems can boost both the communication and sensing (C&S) performance, the classical passive 3D reflect beamforming gain may be negligible due to the ‘double fading’ effect of transmitter-IRS and IRS-receiver links. On the contrast, the active IRS is effective in mitigating such multiplicative fading but by consuming extra energy to amplify the reflecting signals. To alleviate the contradiction between the energy limitation and active beamforming, wireless powered IRS (WP-IRS) is involved into ISAC systems for its self-sustaining ability and easy-to-deploy nature with extremely low power consumption. The paper proposes a matching game, convex optimization and multi-agent deep reinforcement learning (MADRL) combined alternative optimization (AO) algorithm to maximize the sensing beampattern gain towards the direction of targets while satisfying the communication requirement of mobile users. In specific, the sensing targets assignment is transformed into a matching game. The transmission precoding design of BS in two phases can be solved by the majorization-minimization (MM) technique and Lagrangian dual transformation. The cooperative reflecting matrix optimization of multi-IRS is achieved by the value decomposed critic (VDC) augmented MADRL approach, named as MAVDC, which helps to avoid the local sub-optimal by computing the contribution of each IRS and guiding a more robust policy updating. Numerical results reveal that the proposed Match-MM-MAVDC algorithm is effective in speeding up the networks convergence and pursuing a near closed-form solution with low-complexity.