Distributed Antenna Selection and Beamforming for ELAA Based Downlink ISAC Systems

An extremely large-scale antenna array (ELAA) based integrated sensing and communication (ISAC) system is investigated. Focusing on a practical user-roaming environment, the ISAC beamforming design and antenna selection for ELAA are studied. In particular, two antenna selection schemes, namely, movable subarray selection (MSS) and adjustable antenna selection (ASS), are proposed to provide low complexity and flexible beamforming design. For each proposed scheme, a long-term sum rate maximization problem is formulated via jointly optimizing the antenna selection and beamforming, subject to the constraints of transmitting power and beampattern gain for target sensing. To address this problem, a dimension reduction assisted multi-agent proximal policy optimization (DR-MAPPO) algorithm, which can be used for different antenna selection schemes, is developed. An OrderConv layer is designed, which reduces the input dimension by removing the channel state information (CSI) of unselected antennas while keeping the order information of selected ones. Moreover, multiple agents are employed to execute antenna selection and beamforming in a distributed manner according to current channel state information and previous antenna selection, which can reduce the complexity by decomposing the problem. Numerical results demonstrate that MSS outperforms the fixed subarray selection by providing more degrees of freedom and ASS further surpasses MSS through adopting changeable subarray shape. Moreover, the proposed DR-MAPPO algorithm improves the computational efficiency and achieves preferable data rates compared to original MAPPO algorithm.