A Hierarchical MADRL Framework for Joint Resource Allocation and Trajectory Design in UAV Enabled Multi-Static ISAC Networks

Unmanned aerial vehicle (UAV) enabled integrated sensing and communication (ISAC) networks is a promising paradigm. This letter proposes a hierarchical multi-agent deep reinforcement learning (MADRL) framework to jointly optimize user association, UAV trajectory, and power allocation in UAV-ISAC networks. Specifically, a multi-agent deep Q-network (MADQN) is proposed to handle discrete user association decisions, while a multi-agent proximal policy optimization (MAPPO) algorithm is designed to manage continuous trajectory and power control. To enhance long-term decision-making, a long short-term memory (LSTM) layer is incorporated to capture temporal dependencies in UAV mobility and environmental states. Simulation results demonstrate that the proposed approach significantly improves both communication throughput and sensing performance compared to other methods.