Model-based Validation as Probabilistic Inference

TL;DR

This paper introduces a probabilistic inference framework for estimating failure distributions in autonomous systems, leveraging Bayesian methods and automatic differentiation to improve sampling efficiency and coverage.

Contribution

It presents a novel model-based approach that frames failure trajectory estimation as Bayesian inference, enabling more comprehensive and efficient failure analysis.

Findings

Improved sample efficiency over baseline methods

Enhanced coverage of failure modes in complex systems

Demonstrated on diverse autonomous system scenarios

Abstract

Estimating the distribution over failures is a key step in validating autonomous systems. Existing approaches focus on finding failures for a small range of initial conditions or make restrictive assumptions about the properties of the system under test. We frame estimating the distribution over failure trajectories for sequential systems as Bayesian inference. Our model-based approach represents the distribution over failure trajectories using rollouts of system dynamics and computes trajectory gradients using automatic differentiation. Our approach is demonstrated in an inverted pendulum control system, an autonomous vehicle driving scenario, and a partially observable lunar lander. Sampling is performed using an off-the-shelf implementation of Hamiltonian Monte Carlo with multiple chains to capture multimodality and gradient smoothing for safe trajectories. In all experiments, we observed improvements in sample efficiency and parameter space coverage compared to black-box baseline approaches. This work is open sourced.