Research On Autonomous Driving Control Of Four-wheel Independently Driven Articulated Vehicles

With the development of new energy vehicle technology,such as electric driven,four-wheel independently driven articulated vehicles have gradually become important transportation equipment in mining,agriculture,military,and other industries.Therefore,the autonomous driving control of such vehicles has become a core technology of the automation process for these industries.The autonomous driving control of four-wheel independently driven articulated vehicles currently faces two challenges.One is that the traditional path tracking control method performs poorly when the curvature of the reference path sharply changes by a large margin.Another is that the four-wheel independently driven system for articulated vehicles has problems such as uncoordinated driving force distribution and inefficient use of differential torque.To improve the accuracy of the autonomous driving control of the four-wheel independently driven articulated vehicle and reduce the errors in displacement and heading,this paper studies the above two challenges.The study is divided into three main sections,including the mathematical model and other theoretical bases,the path tracking control method at the kinematics level,and the autonomous driving control system at the dynamic level.In the study of mathematical models,firstly,the current researches of kinematics models of articulated vehicles are reviewed.A variety of kinematics models of articulated vehicles are introduced,analyzed,and compared.Finally,it is clear that the classical kinematics model of articulated vehicles proposed by Croke is most suitable for the research of path tracking control methods.In terms of dynamic models,at present,the other dynamic model researches focus on the analysis of the dynamic output characteristics of the articulated vehicle,while ignoring the nonlinear relationship between the input and output.Therefore,based on the Newton-Eulerian method,the force analysis and theoretical derivation of the four-wheel independently driven articulated vehicle are carried out,and the Magic Formula(MF)is introduced to complement the solution of the dynamic state,thus a four degree of freedom dynamic model of the articulated vehicle for autonomous driving control is established.The research of path tracking control method at kinematics level includes two parts.In the first part,because of the successful application of model predictive control(MPC)in mobile robots and other fields,firstly,based on linear model prediction control(LMPC),nonlinear model predictive control(NMPC)and nonlinear error model predictive control(NEMPC),path tracking controllers for articulated vehicles are proposed.Then,these three controllers were compared with the published articulated vehicle path tracking controller based on linear error model predictive control(LEMPC).In the simulation results,if the curve radius of the reference path is 15m,and the reference velocity is 4m/s,the errors of LMPC is divergent,the maximum absolute lateral error and the maximum absolute heading error of LEMPC are 1.862m and 0.206rad,respectively,and the maximum absolute lateral error and the maximum absolute heading error of NEMPC are 2.105m and 0.201rad,respectively,while the maximum absolute lateral error and the maximum absolute heading error of NMPC are only 0.234m and 0.073rad.According to the simulation results,the NMPC performs best among these MPC-based path tracking control methods.In the second part,because the traditional NMPC path tracking control uses the coordinate error between the predicted pose of the articulated vehicle and the tracking target point on the reference path as the optimization target.Therefore,as the reference velocity or the sharp curvature change amplitude of the path increases,the error of the path tracking control also increases.To solve this problem,a path-tracking control method named Point-to-line NMPC(PLNMPC)is proposed by improving the optimization objective target.Moreover,in order to solve the problem of the poor real-time performance of the PLNMPC controller,a PLNMPC Neural Network(PLNMPC-NN)path tracking controller is proposed which uses the PLNMPC controller as a learning sample.According to the simulation results,if the curve radius of the reference path is 10m and the reference velocity is 5m/s,the maximum absolute lateral error and the maximum absolute heading error of the PLNMPC controller are 0.256m and 0.097rad,respectively.The accuracy of the PLNMPC-NN controller is similar to that of the PLNMPC controller,and the maximum calculation time in each control period is only 0.019s,which can meet the real-time requirements of path tracking control.In the study of dynamic level,considering the difference and connection between the input of the kinematics model and the output of the dynamic model,a hierarchical autonomous driving control framework including a path tracking layer and a driving force distribution layer is proposed.Based on this,an autonomous driving control system composed of an equal driving force distribution controller and path tracking controller is proposed.After that,considering that the equal driving force distribution controller ignored the differential torque that can be provided between the wheels,a new NMPC-based driving force distribution controller is proposed and a new autonomous driving control system is built.The simulation results show that the above-mentioned autonomous driving control system can control the four-wheel independent driving articulated vehicle tracking reference path.Compared with the equal driving force distribution controller,the NMPC driving force distribution controller can further improve the accuracy of the autonomous driving control system.If the curve radius of the reference path is 10m and the reference velocity is 4m/s,the maximum absolute lateral error and the maximum absolute heading error of the autonomous driving control system composed of the PLNMPC path tracking controller and the NMPC driving force distribution controller are 0.348m and 0.098rad,respectively.After completing the research work in the above three sections,the autonomous driving control system based on PLNMPC-NN path tracking is tested through an experiment platform which is a 35t independently driven articulated truck developed by the University of Science and Technology Beijing and Shandong Gold Engineering Company.And According to the experiment results,the proposed autonomous driving control system can control the 35t independently driven the articulated truck to track the reference path.If the curve radius of the reference path is 15m,the reference velocity is 4m/s,the maximum absolute lateral error of the autonomous driving control system is 0.289m,and the maximum absolute heading error is 0.074rad.It can be known from these results that the proposed autonomous driving control system has high accuracy when tracking a reference path with a large sharp change in curvature.To sum up,the results of this paper have to some extent solved the problems,such as the difficulty in tracking a reference path with a large sharp change in curvature,which are encountered in the autonomous driving control of four-wheel independent driving articulated vehicles.