基于虚拟样机的微缩智能车仿真教学实验开发

Development and Application of Miniature Smart Car Simulation Teaching Experiment Based on Virtual Prototype

基于虚拟样机的微缩智能车仿真教学实验采用Unity3D技术,根据第十四届全国大学生智能汽车竞赛官方规则设计微缩智能车虚拟样机和虚拟赛道,搭建虚拟比赛环境。通过行为克隆方式采集运行中的驾驶行为数据,进而采用NVIDIA端到端自动驾驶卷积神经网络模型结构进行端到端训练,训练出的模型能够完成较为复杂的赛道任务。本实验疫情期间用于自动化专业选修课系统仿真和全校创业公选课自动驾驶仿真探索,取得良好的教学效果。

The virtual prototype-based miniature smart car simulation teaching experiment developed in this paper uses Unity-3D technology to design the miniature smart car virtual prototype and virtual track according to the official rules of the 14th National University Student Smart Car Competition,build a virtual competition environment,and collect and run through behavioral cloning the virtual track data is then used for end-to-end training using the NVIDIA end-to-end autopilot convolutional neural network model structure.The trained model can complete more complex track tasks.During the epidemic,this experiment was used for the elective course"System Simulation"for automation majors and the public elective course Exploration of Autonomous Driving Simulation for entrepreneurship,and achieved good teaching results.