Multi-Tier UAV Swarm Deployment for JRC Systems: Distributed Optimization With Learning-Based Adaptation

Abstract

This paper proposes a multi-tier coordination framework for autonomous Unmanned Aerial Vehicle (UAV) swarm deployment in joint radar-communication (JRC)-enabled post-disaster assessment. The proposed framework adopts a distributed/coordinated optimization approach, where analytical updates are derived for the initial 3D UAV positioning and bandwidth–power allocation, jointly optimizing sensing quality and communication performance under resource constraints. Building on this foundation, a Deep Reinforcement Learning (DRL) agent is also developed to dynamically refine UAVs’ positions in real time, adapting to environmental uncertainties and mission dynamics. The proposed hybrid optimization–learning framework targets the balance between optimality and complexity by enabling adaptive and intelligent decision-making for real-time and efficient deployment of distributed UAV swarms. The DRL policy adapts well to dynamic, mobile-target scenarios, while the distributed optimization enables rapid and pre-training-free deployment, making it ideal for time-critical missions. Unlike existing approaches that either rely on centralized control or neglect the interplay between sensing and communication, our framework enables distributed, infrastructure-free coordination. Simulation results show that the proposed framework achieves up to 13% and 27% higher average sensing SNR when varying the number of targets and total available power per UAV, respectively, compared to communication-centric, radar-centric, and learning-based baselines. These results confirm the effectiveness of distributed optimization and DRL-based coordination for scalable, resilient, and adaptable UAV deployment in disaster response and other mission-critical scenarios.