Research On Motion Planning And Control Of Unmanned Vehicle In Urban Road Environment

In order to solve the problems such as traffic accidents,road congestion and environmental pollution brought by automobiles to human society,unmanned driving technology has become a hotspot in research at home and abroad,and its core is environmental perception,decisionmaking planning and motion control.Although breakthroughs have been made in the unmanned driving technology in recent years,and some unmanned vehicles have entered the stage of limited-area scenario testing and commercial trial operation,there are still some problems in the above-mentioned core technologies,such as local dynamic motion planning and longitudinal and lateral coupling control,which hinder the application and popularization of unmanned vehicles in real urban road environments.In this paper,the problems in motion planning and control of unmanned vehicle in urban road environments are studied.The main research contents are listed as follows:1.Research on modeling of coding road scene and the creation methond of local dynamic target point model.In order to create the road scene environment required by the motion planning of unmanned vehicles,the local road scene coding model is designed according to the local road topology information and typical driving behaviors recognized by the perception layer.A local dynamic target point model is established based on typical driving behaviors target point model.According to different road topologies,an encoded road scene simulation experiment environment is created.This method can be used to model the road online in real time,and can be used for path optimization by weighting.At the same time,the prediction model of dynamic traffic participant behavior is added to the road model,and combined with the current state transition model,which can be used for later decision-making planning.2.Research on motion planning method of unmanned vehicle based on improved behavior dynamics.Aiming at the problems of slow convergence speed or complex path twisting motion planning method in an environment with obstacles or dense traffic participants,the behavioral dynamics method is introduced into the motion planning of unmanned vehicles,and the classical behavior dynamics model is improved by designing the local dynamic target point model,the coupling relationship between the lateral heading angle and the longitudinal velocity,and the behavior coordination based on particle swarm optimization.The proposed coded road model and the improved behavioral dynamics motion planning model are used test the effectiveness and feasibility in different road scenes,and compared with the polynomial planning method,to verify the smoothness of the behavioral dynamics motion planning trajectory and lay the foundation for motion coupling control.3.Research on motion planning method of behavioral dynamics considering humanvehicle interaction.In view of the problem that the existing motion planning methods cannot effectively describe the two-way interaction of pedestrians-vehicles in the mixed traffic environment,the social force model based on pedestrians-pedestrians interaction is extended,and the virtual force between people-vehicles is expanded and is introduced into the navigation behavior dynamic model,which solved the problem of pedestrians-vehicle two-way dualsystem interaction.The deep learning algorithm is used to detect and track pedestrians and vehicles in the video of the data set in real time,and the above motion planning model is used to simulate in the same scene.Through comparative analysis with actual pedestrian and vehicle motion trajectories,it is proved that the proposed method can effectively carry out motion planning for pedestrians-vehicle interaction scenarios.4.Research on model predictive control strategy based on longitudinal and lateral coupling.Aiming at the safety and comfort problems caused by the independent control of the longitudinal and lateral motion of the existing vehicles,a three-degree-of-freedom vehicle dynamics model is constructed,and the model predictive control(MPC)and an improved neural network PID feedback control method are used to realize the steering,acceleration and deceleration of vehicle.The behavioral dynamics motion planning trajectory is used as the objective function term,a motion tracking control strategy based on model predictive control based on longitudinal and lateral coupling is proposed.Through the tracking control simulation of different motion planning trajectories is simulated,and as well as comparison with independent motion control simulation,the feasibility and effectiveness of the predictive control strategy of the v longitudinal and lateral coupling model are proved.