Diffusion-Based Failure Sampling for Evaluating Safety-Critical Autonomous Systems

TL;DR

This paper introduces a diffusion-based method for efficiently sampling failure scenarios in high-dimensional autonomous systems, significantly improving over traditional black-box approaches in robotic safety validation.

Contribution

The paper presents a novel application of conditional denoising diffusion models to failure sampling, enhancing sample efficiency and mode coverage in safety-critical autonomous system validation.

Findings

Improved sample efficiency over existing methods

Enhanced mode coverage in failure distribution sampling

Effective validation on high-dimensional robotic tasks

Abstract

Validating safety-critical autonomous systems in high-dimensional domains such as robotics presents a significant challenge. Existing black-box approaches based on Markov chain Monte Carlo may require an enormous number of samples, while methods based on importance sampling often rely on simple parametric families that may struggle to represent the distribution over failures. We propose to sample the distribution over failures using a conditional denoising diffusion model, which has shown success in complex high-dimensional problems such as robotic task planning. We iteratively train a diffusion model to produce state trajectories closer to failure. We demonstrate the effectiveness of our approach on high-dimensional robotic validation tasks, improving sample efficiency and mode coverage compared to existing black-box techniques.