2026/06/30 by Weixian Qian, Linfeng Liang, Tianyi Yang +7
Computer Science · Engineering · #Adversarial Robustness in Machine Learning #Autonomous Vehicle Technology and Safety #Descent (aeronautics) #Diversity (politics) #Gradient descent #Hazardous waste #Seeding #Software Reliability and Analysis Research #Variety (cybernetics)
paper · pdf · open access · doi:10.1145/3808153
published in Proceedings of the ACM on software engineering. 3(FSE), 3298-3320 (Association for Computing Machinery)
openalex created_date 2026/04/24 · openalex publication_date 2026/06/30 · openalex updated_date 2026/08/05
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.