汽车工程学报2025，Vol.15Issue(3)：329-339,11.DOI:10.3969/j.issn.2095-1469.2025.03.06
基于大语言模型的自动驾驶仿真测试场景生成
Simulation Test Scenario Generation for Autonomous Driving Based on Large Language Models
摘要
Abstract
The rapid development of autonomous driving technology has increased the demand for the authentic and diverse simulation test scenarios.However,traditional methods for constructing autonomous driving simulation scenarios heavily rely on manual editing,which is not only costly but also limited by the combination and complexity of scene elements,making it difficult to meet the comprehensive testing and validation needs of autonomous driving systems.To address this issue,this paper proposes a method for generating autonomous driving simulation test scenarios based on Large Language Models(LLMs).This approach utilizes a pre-trained LLM,enhanced through LoRA fine-tuning,and integrates a scenario language parser to produce a structured interpretive language,which is used to generate scenario files.The generated text is processed by a parser to convert it into usable scenario files,effectively addressing the issues of overly long texts and model hallucinations,while also achieving the specialization of a general model s capabilities.关键词
人工智能/自动驾驶/场景生成大模型/仿真Key words
artificial intelligence/autonomous driving/scenario generation model/simulation分类
计算机与自动化引用本文复制引用
陈贞,李京泰,郭煌,贾贝贝,李广友..基于大语言模型的自动驾驶仿真测试场景生成[J].汽车工程学报,2025,15(3):329-339,11.基金项目
新一代人工智能国家科技重大专项(2022ZD0116311) （2022ZD0116311）
