PCASim: Promptable Closed-loop Adversarial Simulation for Urban Traffic Environment

2026-05-15

AI Summary

This study addresses the lack of an efficient closed-loop safety testing framework for urban autonomous driving, which hinders the co-evolution of adversarial scenario generation and agent safety optimization. The authors propose a novel closed-loop framework integrating large language models (LLMs), a knowledge base, and reinforcement learning. In this framework, a promptable LLM generates high-fidelity, safety-critical edge-case traffic scenarios on demand; rule-based filtering and knowledge retrieval ensure semantic plausibility; and diverse vehicle agents are jointly trained within a closed-loop simulation environment. This approach achieves, for the first time, synergistic co-evolution between adversarial scenario generation and safety-oriented agent training, yielding a 12% improvement in domain-specific language generation accuracy, an 8% increase in successful transfer to novel scenarios, and a 30% enhancement in obstacle avoidance capability.

Abstract

Real-world autonomous driving, particularly in urban environments with numerous corner cases, requires rigorous testing to ensure product safety and robustness. However, few studies have explored integrating adversarial scenario generation with the training of safety agents in closed-loop testing, enabling efficient co-evolution and mutual enhancement of both. To address this challenge, an adversarial behavior knowledge repository is constructed by applying rule-based filtering to an open-source dataset, combined with knowledge retrieval modules tailored for simulation environments. A large language model (LLM) is employed to integrate knowledge-, data-, and adversarial-driven approaches, generating safety-critical traffic scenarios customized to user needs. Additionally, while evaluating the generated scenarios, we employ reinforcement learning models to train the behaviors of different types of vehicles, thereby enriching scenario diversity beyond existing datasets while preserving realism. Experimental results demonstrate that the proposed framework improves the accuracy of domain-specific language generation by 12\%. Moreover, the success rate of newly generated scenario transformations increases by 8\%, while obstacle-avoidance capability is enhanced by 30\%. For the complete manuscript, please refer to: https://zhenhaooo.github.io/PCASim.github.io/

Problem

• adversarial simulation
• closed-loop testing
• urban traffic environment
• autonomous driving
• safety-critical scenarios

Innovation

• Promptable Simulation
• Closed-loop Adversarial Testing
• Large Language Model (LLM)
• Reinforcement Learning for Behavior Modeling
• Safety-critical Scenario Generation