Meta-Learning-Driven Resource Optimization in Full-Duplex ISAC with Movable Antennas

This paper investigates a full-duplex (FD) scenario where a base station (BS) equipped with movable antennas (MAs) simultaneously provides communication services to a set of downlink (DL) and uplink (UL) users while also enabling sensing functionalities for target detection, thereby supporting integrated sensing and communication (ISAC) technology. Additionally, a receiving BS, also equipped with MAs (denoted as BS R), is responsible for capturing the reflected echo. To optimize this setup, we formulate an optimization problem aimed at maximizing the signal-to-noise and interference ratio (SINR) of the captured echo. This is achieved by jointly optimizing the transmit beamforming vectors at the FD BS, the receiving beamforming vectors at both the FD BS and BS R, the UL users' transmit power, and the MAs' positions at both BSs, all while satisfying the quality-of-service (QoS) requirements for both sensing and communication. Given the non-convex nature of the problem and the high coupling between the variables, we employ a gradient-based meta-learning (GML) approach tailored for large-scale optimization. Numerical results demonstrate the effectiveness of the proposed meta-learning approach, achieving results within 99% of the optimal solution. Furthermore, the MA-based scheme outperforms several benchmark approaches, highlighting its advantages in practical ISAC applications.