From Threat Intelligence to Safe Paths: Cooperative UAV–UGV Navigation in Contested Urban Search and Rescue

Autonomous Ground Vehicle (UGV) navigation in contested urban search-and-rescue (SAR) settings must balance mission completion against exposure to mobile adversaries whose positions are known only from aging, intermittent observations. This work presents a cooperative UAV-UGV navigation framework in which an aerial agent supplies threat intelligence to a ground agent running a risk-aware A* path planner over a real urban road graph. The core contribution is a temporal decay threat model that attenuates each adversary detection's influence as a continuous function of observation age, combining exponential confidence decay with a worst-case reachability-expanding uncertainty radius, integrated directly into the planner's edge-cost function. Five UAV coverage strategies, no UAV, lawnmower, threat-informed, target-informed, and a novel phase-scheduled hybrid, were evaluated across 300 simulated missions on an OpenStreetMap-derived road network, scored by a composite Mission Performance Index (mission success, adversary containment, time efficiency). At standard adversary density, threat-informed aerial coverage achieves the highest UGV success rate. At triple density, no UAV-assisted strategy outperforms the baseline on the composite index, revealing a critical density threshold, bounded empirically between the two tested densities, beyond which the UAV coordination overhead becomes net-negative for ground-agent performance. The results carry direct design implications for density-conditional UAV deployment policies in multi-tier autonomous SAR systems.