TrafficAlign: Aligning Large Language Models for Traffic Scenario Generation
For autonomous driving safety, TrafficAlign provides a method to generate realistic traffic scenarios that improve model robustness, though it is an incremental improvement over existing LLM-based generation approaches.
TrafficAlign aligns LLMs to generate realistic traffic scenarios from real-world driving videos, outperforming state-of-the-art methods by revealing up to 10.8% more collisions and reducing collision rates by 36.1% when used for fine-tuning.
Recent research has investigated the use of large language models (LLMs) to generate traffic scenarios for autonomous driving. However, pretrained LLMs often fail to align with real-world traffic distributions. In this work, we present TrafficAlign, an automated framework that synthesizes traffic scenarios based on real-world driving videos, performs data validation, and aligns LLMs with the synthesized scenarios. The evaluation shows that traffic scenarios generated by TrafficAlign are highly effective, revealing up to 10.8% more collisions on average across three autonomous driving models than state-of-the-art methods. Furthermore, fine-tuning these driving models with TrafficAlign-generated scenarios significantly reduced collision rates by 36.1% compared with the original models. A qualitative study using traffic datasets from six geographically diverse regions shows that TrafficAlign-generated scenarios exhibit strong alignment with corresponding traffic distributions in these regions.