摘要
针对现有智能汽车基于场景测试方法严重依赖人力、效率瓶颈凸显的问题,本文提出了一种基于大语言模型的智能汽车仿真测试方法。首先,设计基于大语言模型的智能汽车仿真测试架构,建立了对应的数据层和仿真层;在此基础上,构建了基于大语言模型的智能汽车仿真测试流程,针对知识问答型任务设计了知识挖掘、模型微调与知识库增强检索应用流程,针对场景生成任务设计了场景类型分析、场景要素生成、场景工具链调用的应用路径,针对测试评价型任务,设计了测试场景解析、评价体系构建与仿真测试执行综合应用框架;最后,对各任务进行了测试。结果证明,本文所提出的测试方法可以有效解决不同类型的测试任务,提升测试效率。
In this paper a simulation testing method for intelligent vehicle based on a large language model is proposed to address the issues of heavy reliance on human resources and prominent efficiency bottlenecks in existing scenario based testing methods.Firstly,a simulation testing architecture for intelligent vehicle based on a large language model is designed,and corresponding data and simulation layers are established.On this basis,an intelligent car simulation testing process based on a large language model is constructed.Knowledge mining,model finetuning,and knowledge base enhancement retrieval application processes are designed for knowledge question answering tasks.Application paths for scenario type analysis,scenario element generation,and scenario toolchain invocation are designed for scenario generation tasks.For testing and evaluation tasks,a comprehensive application framework for testing scenario analysis,evaluation system construction,and simulation testing execution is designed.Finally,each task is tested.The results show that the testing method proposed in this paper can effectively solve different types of testing tasks and improve testing efficiency.
作者
朱冰
汤瑞
赵健
张培兴
李文旭
李嘉胜
徐雪峰
Zhu Bing;Tang Rui;Zhao Jian;Zhang Peixing;Li Wenxu;Li Jiasheng;Xu Xuefeng(Jilin University,National Key Laboratory of Automotive Chassis Integration and Bionics,Changchun 130022)
出处
《汽车工程》
北大核心
2025年第4期587-597,共11页
Automotive Engineering
基金
国家自然科学基金(U22A20247,52172386)
中国博士后科学基金(2023M741354,GZC20230945)资助。
关键词
智能汽车
仿真测试
大语言模型
场景生成
自动测试
intelligent vehicle
simulation testing
large language model
scenario generation
automatic testing
