Research papers · Posted 2025.08.20 · Preprint

Adversarial Generation and Collaborative Evolution of Safety-Critical Scenarios for Autonomous Vehicles

Jiangfan Liu, Yongkang Guo, Fangzhi Zhong, Tianyuan Zhang, Zonglei Jing, Siyuan Liang, Jiakai Wang, Mingchuan Zhang, Aishan Liu, Xianglong Liu · arXiv · English

Automated driving Scenarios

Overview

ScenGE uses an LLM to formulate plausible adversarial behaviors, then evolves background traffic to amplify threats and occlusions. Simulation, vehicle tests and human evaluation examine scenario criticality, plausibility and robustness-training benefits.

Version: v2

Adversarial Generation and Collaborative Evolution of Safety-Critical Scenarios for Autonomous Vehicles

Overview

Participation enquiry

Your enquiry has been sent.

Adversarial Generation and Collaborative Evolution of Safety-Critical Scenarios for Autonomous Vehicles

Overview

Related resources

Assuring Learning-Enabled Cyber-Physical Systems with Foundation Models: A Neuro-Symbolic Agenda for Testing, Verification, and Runtime Assurance

Expand-and-Prune: Test-Time Candidate-Space Adaptation for End-to-End Autonomous Driving

ProbGuard: Proactive Runtime Monitoring for LLM Agent Safety via Probabilistic Prediction