摘要
随着自动驾驶技术的不断发展,传统的真实路测方法已经不能满足自动驾驶测试的需求,无法在充分性、全面性、效率性等各个方面达到要求。利用自动驾驶仿真测试场景库进行测试可以大大缩短测试时间和成本,可以提高测试的安全性和可靠性,是助力自动驾驶领域发展的重要工具,已成为目前自动驾驶研究领域的一个重大研究方向。本文在调研分析大量自动驾驶仿真测试场景库研究成果的基础上,首先针对场景作为自动驾驶仿真测试场景库的主体进行一系列的分析,梳理了场景的定义、场景的元素、场景的数据来源;其次,总结了常见场景数据处理,主要包括数据预处理、场景特征提取、场景聚类分析;再次,归纳了自动驾驶仿真测试场景库体系的搭建、分布式架构、基于场景库的V字开发模型;然后,在此基础上提出了基于场景库开发的应用;最后,对目前自动驾驶仿真测试场景库的研究提出了一些未来的研究展望。本文对自动驾驶仿真测试场景库构建的研究可以给自动驾驶测试开发研究者提供一定的参考,有助于推动自动驾驶技术的发展。
With the continuous development of autonomous driving technology,traditional real road testing methods can no longer meet the needs of autonomous driving testing,and cannot achieve sufficiency,comprehensiveness,efficiency and other aspects.Using the automated driving simulation test scenario library for testing can greatly shorten the test time and cost,and improve the safety and re⁃liability of the test.It is an important tool to help the development of the autonomous driving field and has become a major research di⁃rection in the current autonomous driving research field.Based on the investigation and analysis of the research results of a large num⁃ber of automated driving simulation test scene libraries,this article first conducts a series of analysis on the scene as the main body of the automated driving simulation test scene library,and sorts out the definition of the scene,the elements of the scene,and the data source of the scene;Secondly,it summarizes common scene data processing,mainly including data preprocessing,scene feature extrac⁃tion,scene clustering analysis;Thirdly,it summarizes the construction of the automatic driving simulation test scene library system,dis⁃tributed architecture,and the V word development model based on the scene library;From the second,on this basis,the application based on the development of the scene library is proposed;Finally,some future research prospects are put forward for the current re⁃search on the automatic driving simulation test scene library.The research in this article on the construction of the automated driving simulation test scenario library can provide a certain reference for the researcher of automated driving test development and help pro⁃mote the development of autonomous driving technology.
作者
曾德松
高琛
谭北海
余荣
Zeng Desong;Gao Chen;Tan Beihai;Yu Rong(Guangdong University of Technology,Guangzhou 510006;Guangzhou Wei Chi Technology Co.,Ltd.,Guangzhou 511455)
出处
《现代计算机》
2021年第31期1-12,32,共13页
Modern Computer
基金
国家自然科学基金(61971148):车联网边缘智能的资源优化与服务协同研究。
关键词
自动驾驶
场景库
仿真测试
场景生成
autonomous driving
scene library
simulation test
scenario generation