Joint Active and Passive Beamforming Design for Intelligent Reflecting Surface Assisted Integrated Sensing and Communications Systems

As a promising emerging technology, the combination of an integrated sensing and communications (ISAC) system and intelligent reflecting surface (IRS) can obtain better comprehensive properties for both communication and target sensing. To explore the potential ability of a multi-IRS assisted ISAC system, both the active beamforming matrix and passive beamforming matrix should be delicately jointly optimized for simultaneous communication and sensing enhancement, which introduce additional hardware complexity and difficulty to obtain a solution by the conventional methods. Therefore, this paper proposes a successive convex approximation (SCA) and multi-agent deep reinforcement learning (MADRL) combined approach to enhance the communication capability with the target sensing beampattern constraints. In particular, numerical results show that the sensing performance can be guaranteed with the boosting of communication capability, especially when the users are not directly visible to the BS.