Runtime Safety Filtering for Learned Small UAS Separation Policies under GNSS Degradation
2026-07-10 • Robotics
RoboticsMachine LearningMultiagent Systems
AI summaryⓘ
The authors studied how small drones can avoid crashes when their GPS signals are unreliable, such as in cities where signals get blocked or jammed. They compared two methods: one that changes the drone’s actions after the drone decides, and another that fixes the drone’s input data before it decides what to do. Their tests showed that fixing the input data works much better, reducing near collisions by 90%, while changing actions after the decision helped very little. This means letting the drone’s own system make decisions based on better information is safer than overriding its actions later.
small Unmanned Aircraft Systems (sUAS)separation assuranceGlobal Navigation Satellite Systems (GNSS)multipath propagationcontrol barrier functionsaction filteringobservation filteringcollision avoidancemulti-agent systemsnavigation integrity
Authors
Alex Zongo, Peng Wei
Abstract
Learning-based separation assurance for small Unmanned Aircraft Systems (sUAS) achieves near-zero collision rates in simulation, but assumes accurate position and velocity information from Global Navigation Satellite Systems (GNSS). This assumption fails in urban environments, where multipath propagation, signal blockage, and intentional interference degrade navigation integrity. This raises a fundamental architectural question for deploying learned separation policies under GNSS degradation: should runtime safety mechanisms filter the policy's actions or its observations? This work evaluates both approaches for multi-agent sUAS separation under adversarial GNSS degradation. Both architectures first estimate a worst-case traffic state consistent with bounded observation uncertainty, then diverge: action filtering constrains policy outputs via discrete-time control barrier functions evaluated at the worst-case state, while observation filtering presents the worst-case state directly to the policy as corrected input. Experimental results show that action filtering provides negligible safety improvement, while observation filtering reduces near mid-air collisions by 90% and remains robust to the barrier function's tradeoff between separation distance and closing rate. These results suggest that, for policies with learned safety behaviors, preserving the policy's decision authority outperforms overriding its actions with hand-designed constraints.