多控制模型嵌入的智能汽车防撞虚拟仿真

Collision avoidance virtual simulation of intelligent vehicle embedded with multiple control models

针对智能汽车行驶安全距离监测与防撞试验高成本、高危险性以及试验结果难以观察的问题,研究了智能汽车安全距离监测与防撞的虚拟仿真;应用Visual Studio 2015、3Dmax、Unity3D等虚拟仿真技术在虚拟制动控制系统中进行了可嵌入多控制模型的自动驾驶汽车行车安全距离监测和防撞虚拟仿真试验,测试了不同制动模型的防撞应用效果;以电动汽车动力性能和制动力学特征为基础建立了虚拟整车模型及其制动系统模型,根据路面附着系数和不同道路材质建立了虚拟试验道路模型和试验场景等虚拟环境,开发了试验电子设备仿真模型,实现了多虚拟控制器嵌入,研究了基础模型和反向传播(BP)神经网络模型的嵌入与仿真效果;通过设计接口将虚拟软硬件设计效果与仿真试验过程相关联,并采用动画渲染直观展现,同时应用内存优化实现了在网络版访问服务器中进行虚拟仿真试验;通过实车试验验证了仿真系统,并对比了实车试验数据与2种模型的仿真数据。研究结果表明:在低速情况下,基础模型计算的安全距离与实车试验所测安全距离的相对误差为2.49%,BP神经网络模型预测的安全距离与实车试验所测安全距离的相对误差为2.07%;在高速情况下,因为传感器不稳定的原因,基础模型计算的安全距离与实车试验所测安全距离的相对误差为10.03%,BP神经网络模型预测的安全距离与实车试验所测安全距离的相对误差为10.35%。由此可见,该仿真系统可使高风险的碰撞试验在虚拟环境下完成。

In view of the high cost and risk involved in the driving safety distance monitoring and collision avoidance tests of intelligent vehicles and the difficulty in visualizing the test results,virtual simulations were conducted for the safety distance monitoring and collision avoidance of an intelligent vehicle.The virtual simulation technology based on the Visual Studio 2015,3 Dmax,and Unity3 D was used to conduct the driving safety distance monitoring and collision avoidance virtual simulation tests on an autonomous car equipped with multiple control models in the virtual braking control system.The collision avoidance effects of different braking models were tested.Models for a virtual full vehicle and its braking system were established based on the dynamic performance and braking dynamics characteristics of electric vehicles.A virtual environment including a virtual test road model and a test scene was established based on the pavement adhesion coefficient and different road materials.Simulation models were developed for the test electronic devices to realize the multiple virtual controllers embedding.The embedding and simulation effects of the basic model and a back propagation(BP)neural network were studied.The design effects of the virtual software and hardware were correlated with the simulation test process through the design interface,and the animation rendering was used to directly display visually.Meanwhile,the memory optimization was applied to enable the virtual simulation test to be conducted in the server through the web access.Actual vehicle tests were conducted to verify the simulation system,and the actual vehicle test data were compared with the simulation results using the two models.Research results indicate that,at a low velocity,the relative error between the safety distance calculated by the basic model and that measured by the actual vehicle test is 2.49%,while the relative error between the safety distance calculated by the BP neural network and that measured by the actual vehicle test is 2.07%.At a high velocity,because of the sensor instability,the relative error between the safety distance calculated by the basic model and that measured in the actual vehicle test is 10.03%,while the relative error between the safety distance calculated by the BP neural network and that measured in the actual vehicle test is 10.35%.Therefore,the simulation system can enable a high-risk collision test to be conducted in a virtual environment.6 tabs,16 figs,32 refs.