From Particles to Perils: SVGD-Based Hazardous Scenario Generation for Autonomous Driving Systems Testing

TL;DR

This paper introduces PtoP, a novel framework using SVGD for generating diverse, failure-inducing scenarios in autonomous driving system testing, outperforming existing methods in safety violations and scenario diversity.

Contribution

PtoP combines adaptive seed generation with SVGD to improve the diversity and effectiveness of failure scenario generation in autonomous driving testing.

Findings

PtoP increases safety violation detection rate by up to 27.68%.

It enhances scenario diversity by 9.6%.

It improves map coverage by 16.78%.

Abstract

Simulation-based testing of autonomous driving systems (ADS) must uncover realistic and diverse failures in dense, heterogeneous traffic. However, existing search-based seeding methods (e.g., genetic algorithms) struggle in high-dimensional spaces, often collapsing to limited modes and missing many failure scenarios. We present PtoP, a framework that combines adaptive random seed generation with Stein Variational Gradient Descent (SVGD) to produce diverse, failure-inducing initial conditions. SVGD balances attraction toward high-risk regions and repulsion among particles, yielding risk-seeking yet well-distributed seeds across multiple failure modes. PtoP is plug-and-play and enhances existing online testing methods (e.g., reinforcement learning--based testers) by providing principled seeds. Evaluation in CARLA on two industry-grade ADS (Apollo, Autoware) and a native end-to-end system shows that PtoP improves safety violation rate (up to 27.68%), scenario diversity (9.6%), and map coverage (16.78%) over baselines.