A Reinforcement Learning-Driven STAR-RIS-Assisted UAV for Integrated Sensing, Communication, and Computation

In this paper, we present an integrated sensing, communication, and computing (ISCC) framework tailored for unmanned aerial vehicle (UAV) systems, aimed at preprocessing radar-sensed data before offloading it to the base station (BS) for computation. To enhance full-space coverage capabilities, we use the simultaneously transmitting and reflecting surface reconfigurable intelligent surfaces (STAR-RIS) technology, which empowers the ISCC network by enabling dynamic control over signal transmission and reflection. The proposed framework achieves optimal weighted sum-rate performance while satisfying fundamental sensing requirements through joint optimization of ISCC beamforming, subcarrier allocation, and STAR-RIS-based transmission and reflection beamforming. The Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm is implemented and evaluated against baseline schemes, demonstrating improved performance.