B. Xia, I. Mantegh and M. Bolic, “Spatiotemporal Barrier and Gap-Guided APF for Deadlock-Resilient UAV Collision Avoidance,” IEEE CCTA 2026.

This paper develops a progressive artificial potential field (APF) framework for UAV navigation in environments containing static obstacles and moving intruders. The baseline method uses LiDAR filtering and direction-dependent repulsion to encourage lateral and vertical avoidance while maintaining progress toward the goal. A second variant incorporates sector-wise time-to-collision estimates to strengthen responses to approaching threats, together with barrier-style repulsion near a safety boundary. The final variant adds a histogram-based gap selector that identifies open directions and introduces an additional guidance force to overcome local minima caused by opposing attractive and repulsive forces.

The three variants are evaluated through 30 randomized simulation trials per method in a 3D environment containing a narrow corridor, a U-shaped obstacle, and two moving intruder UAVs. The complete gap-guided method achieves 100% mission success, compared with 20% for the baseline and 43.3% for the temporal-barrier variant, demonstrating the benefit of combining early threat response with explicit guidance around obstructing geometry. These improvements require greater control effort and modestly longer trajectories. The findings remain limited to the tested simulations; dense clutter and unreliable LiDAR returns may impair gap selection, and general safety guarantees and real-world validation remain to be established.