Leveraging Communication-Assisted Sensing for Efficient Air-Ground Coordination via Heterogeneous MADRL

The Air-Ground Integrated Sensing and Communication Systems (AG-ISAC) leverages the combined strengths of Unmanned Aerial Vehicles (UAVs) and ground networks to deliver enhanced communication and sensing in complex environments. This paper introduces a novel heterogeneous cooperative Multiagent Deep Reinforcement Learning (MADRL) algorithm for AG-ISAC systems, which optimizes communication-assisted sensing through the adoption of air-ground non-orthogonal multiple access, enabling energy-efficient networked sensing while balancing the cooperation and individuality of heterogeneous vehicles. The proposed algorithm features two primary modules: an individual characteristics optimization module, which identifies and harnesses the unique features of UAVs and Unmanned Ground Vehicles (UGVs) to facilitate wide-area sensing data collection, and a heterogeneous cooperative strategy optimization module, which models UAV-UGV collaboration to optimize trajectories and sensing data acquisition. Simulation results validate that the algorithm significantly enhances sensing efficiency, data integrity, and power management, representing an intelligent solution for integrated air-ground communication and sensing in constrained urban environments.