Dynamic Air–Ground Risk Fields for UAVs in Road Traffic Environments: A Traffic-Aware Predictive Framework

This study addresses the insufficient representation of moving ground targets in conventional UAV air–ground risk assessment, which typically relies on static population distributions or instantaneous target locations and therefore fails to capture the spatiotemporal evolution of risk in dynamic road traffic environments. A traffic-aware predictive framework for constructing dynamic air–ground risk fields is proposed. The framework characterizes dynamic risk through predictive vehicle exposure modeling and spatiotemporal risk-field construction. Specifically, future vehicle trajectories are predicted from observed traffic states and converted into time-dependent occupancy probabilities to represent potential vehicle exposure. A dynamic capsule-shaped hazard region is constructed by coupling UAV failure-induced falling dynamics with UAV–vehicle relative motion, enabling the conflict region to adapt to vehicle motion states. Vehicle occupancy probability and collision severity are then integrated to quantify time-varying air–ground conflict risk. Furthermore, multi-vehicle risks are spatially aggregated and updated over a rolling time horizon to continuously construct the dynamic risk field. Experiments combining real-world vehicle trajectories from the highD dataset with simulated UAV flight data show that the proposed framework captures the formation, merging, redistribution, and contraction of high-risk regions as traffic evolves. These results demonstrate its capability to characterize traffic-induced spatiotemporal risk evolution and provide dynamic risk information for UAV operational risk assessment and risk-aware trajectory planning.