Research On Trajectory Tracking Control For Intelligent Vehicle Based On Neural Network Inverse Method

Intelligent vehicle has become one of the research hotspots in the field of automotive engineering in the world.As the key technology of intelligent vehicle,vehicle motion planning and control method plays a decisive role in whether the intelligent vehicle can drive safely,reasonably and comfortably on the road,which is deeply concerned by researchers at home and abroad.Under the condition of active steering collision avoidance,the target trajectory is complex,and the longitudinal and lateral movements exist simultaneously and are coupled seriously,which puts forward higher requirements for the trajectory tracking control of intelligent vehicles.Based on the subproject of national key research and development program(2017YFB0102603-3),this study takes the four-wheel independent drive intelligent electric vehicle as the research object,takes the active steering collision avoidance of the vehicle with serious longitudinal and lateral coupling as the research scene,and focuses on the three-way motion decoupling tracking control method of the intelligent vehicle,aiming to improve the accuracy of vehicle tracking control.The main research work of this article is as follows:(1)Aiming at the problem that the traditional artificial potential field method doesn’t fully consider the characteristics of structured road and obstacle vehicles in motion planning,the improved risk potential field evaluation model is adopted.By introducing the methods of traffic participant selection and single lane change planning,combined with the speed planning model,the active collision avoidance trajectory planning method of intelligent vehicle which is suitable for the road traffic environment and in line with driving habits of drivers is proposed 。(2)Aiming at the problem that the existing intelligent vehicle trajectory tracking control method cannot track the three plane movements of the vehicle at the same time,the longitudinal and lateral coupling channels interfere with each other to affect the tracking performance,and it is necessary to establish a complex and accurate vehicle model,a longitudinal and lateral decoupling control method is designed based on the advantages of BP neural network with strong self-learning ability and the characteristics of independent of the deep knowledge of the object.Based on the neural network inverse system model of intelligent vehicle,a pseudo linear system is established,which decouples the vehicle plane motion dynamics system with strong coupling characteristics into three independent and simple subsystems,and realizes the decoupled tracking of the three plane motions at the same time.(3)Aiming at the problem that the pseudo linear system can’t track the target track directly,combining the target path,target speed and vehicle motion model,the target state prediction model which is capable of docking vehicle upper track planning module is established.In order to improve the tracking performance of the control system,an additional closed-loop control loop and a target state update loop are designed.(4)In order to verify the effectiveness of the trajectory tracking control algorithm proposed in this paper,a joint simulation platform based on MATLAB / Simulink and Car Sim software is built.The results show that the intelligent vehicle trajectory tracking control algorithm proposed in this paper can accurately track the target trajectory,and has significant advantages in tracking accuracy,motion smoothness and state prediction.