Distributed Multi-UAV MARL for Joint Relay Connectivity and Aerial Sensing in Post-Disaster Networks

In post-disaster scenarios where terrestrial communication infrastructure is partially damaged, a unmanned aerial vehicles (UAV) swarm must jointly maintain end-to-end connectivity while collecting uplink data from ground users and performing aerial imaging over designated target areas. This paper proposes a relay-assisted multi-UAV system model and formulates the joint relaying–uplink reception–target imaging decision-making problem as a multi-objective optimization over connectivity reliability, communication performance, and sensing coverage under UAV mobility constraints. To solve the resulting high-dimensional coupled control problem, we develop a distributed multi-agent reinforcement learning framework in which each UAV learns a decentralized policy from local observations, guided by a carefully shaped reward that enforces connectivity continuity and balances communication–sensing trade-offs. Extensive simulations demonstrate that the proposed approach consistently improves (i) connectivity outage / link continuity, (ii) user data reception throughput (or success rate), and (iii) target-area sensing coverage compared with representative baselines, while exhibiting superior adaptability under dynamic user distribution and link disruptions.