Dynasto: Validity-Aware Dynamic-Static Parameter Optimization for Autonomous Driving Testing

2026-03-22

AI Summary

This work addresses the challenge that existing autonomous driving simulation testing methods struggle to generate traffic scenarios that are both safety-critical and behaviorally plausible, often misclassifying invalid adversarial behaviors as system failures. To overcome this, the authors propose Dynasto, a novel framework that jointly optimizes static initial conditions and dynamic adversarial behaviors for the first time. Dynasto integrates reinforcement learning to train adversarial agents, employs a genetic algorithm to search for critical initial scenarios, and enforces validity constraints grounded in temporal logic and the ISO 34502 safety distance model to ensure realistic failure cases. Furthermore, it leverages graph-based clustering on semantic event sequences to establish an end-to-end pipeline for generating and categorizing meaningful failures. Experiments in HighwayEnv on two controller types demonstrate that Dynasto discovers 60%–70% more valid failures than pure reinforcement learning baselines and identifies approximately twelve interpretable failure modes.

Technology Category

• Search and Optimization: Sampling/Simulation-based Search
• Multiagent Systems: Adversarial Agents
• Machine Learning: Adversarial Learning & Robustness

Application Category

• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Extensive simulation-based testing is important for assuring the safety of autonomous driving systems (ADS). However, generating safety-critical traffic scenarios remains challenging because failures often arise from rare, complex interactions with surrounding vehicles. Existing automatic scenario-generation approaches frequently fail to distinguish genuine ADS faults from collisions caused by implausible or invalid adversarial behaviors, and they typically optimize either scenario initialization or agent behavior in isolation. We propose Dynasto, a two-step testing approach that jointly optimizes initial scenario parameters and dynamic adversarial behaviors to uncover realistic safety-critical failures. First, we train an adversarial agent using reinforcement learning (RL) with temporal-logic-based validity criteria and a safe-distance model inspired by ISO 34502 to promote behaviorally plausible failures. Second, a genetic algorithm (GA) searches over initial conditions while replaying the adversary's failure-inducing behaviors to reveal additional failures that the RL agent alone does not uncover. Finally, a graph-based clustering pipeline groups failures into representative modes based on semantic event sequences. Our evaluation experiments in HighwayEnv across two ADS controllers show that Dynasto finds 60%-70% more valid failures than an RL-only adversary under the same evaluation budget. With clustering, we obtain about 12 interpretable failure modes per system under test, revealing valid failures driven by weaknesses in ego-controller behavior. These results indicate that coordinated dynamic-static optimization with explicit validity constraints is effective for exposing safety-relevant failures in ADS testing.

Problem

• autonomous driving testing
• safety-critical scenarios
• validity-aware failure detection
• scenario generation
• adversarial behavior

Innovation

• dynamic-static optimization
• validity-aware testing
• adversarial reinforcement learning
• genetic algorithm
• failure mode clustering