Neural Bridge Sampling for Evaluating Safety-Critical Autonomous Systems
This addresses the challenge of expensive and unscalable real-world testing for safety-critical domains like autonomous driving and medical robotics.
The paper tackles the problem of evaluating safety-critical autonomous systems by developing a novel rare-event simulation method to compute the probability of dangerous events, demonstrating its efficacy on various scenarios for rapid sensitivity analysis and model comparison.
Learning-based methodologies increasingly find applications in safety-critical domains like autonomous driving and medical robotics. Due to the rare nature of dangerous events, real-world testing is prohibitively expensive and unscalable. In this work, we employ a probabilistic approach to safety evaluation in simulation, where we are concerned with computing the probability of dangerous events. We develop a novel rare-event simulation method that combines exploration, exploitation, and optimization techniques to find failure modes and estimate their rate of occurrence. We provide rigorous guarantees for the performance of our method in terms of both statistical and computational efficiency. Finally, we demonstrate the efficacy of our approach on a variety of scenarios, illustrating its usefulness as a tool for rapid sensitivity analysis and model comparison that are essential to developing and testing safety-critical autonomous systems.