Research On Tracking Control Of Autonomous Vehicle Based On Model Predictive Control

The development of autonomous driving technology provides solutions to traffic safety issues and provides a strong driving force for industrial upgrading.Vehicle tracking control is the connection between the vehicle chassis and the automatic driving system,and it supports the overall algorithm architecture.Model predictive control has the ability to consider predictive information and handle multiple constraints,providing a new approach for vehicle motion control with multiple application scenarios and multiple control targets.Therefore,it is of great significance to study the application of MPC in vehicle tracking control,to improve the tracking effect,driving stability and efficiency of solution under multiple control targets.A vehicle model and road model has been constructed suitable for planning and tracking control.Based on the Frenet-Serret formula,a frenet parameterization method of boundary type road is designed by combining with the frenet parameterization method of curvilinear path,which is convenient to obtain the relationship between center line and boundary of reference path.Considering the change of road adhesion coefficient,a lateral and longitudinal dynamic vehicle model based on the Brush tire model is provided.For target paths in curvilinear form,a MPC path tracking lateral controller(KNMPC)based on approximate tracking error is proposed.By designing an approximate tracking error in the vehicle coordinate system,the 2-level iteration problem caused by searching for the closest point is avoided.An adaptive correction mechanism is proposed to solve the fitting problem under the path of large curvature,which realizes the direct tracking control of the continuous path,and gets rid of the dependence on the external track point generation module.The simulation results show that the proposed method can effectively reduce the tracking error and improve the speed of calculation.Moreover,the algorithm is further verified in the engineering experiment,which shows good tracking ability and real-time performance,and can meet the control requirements of autonomous vehicles in urban conditions.In view of the high complexity of nonlinear MPC,a linear time-varying path tracking control method(DLMPC)based on piecewise linear approximation error is proposed.By combineing the piecewise linearization of the proposed approximate tracking error with the linear discrete dynamic vehicle model,a quadratic form path tracking optimization problem is constructed,which realizes the linear simplification of proposed KNMPC.The simulation results show that the proposed method not only improves the tracking effect under high-speed conditions,but also enhances the solution efficiency and stability.For target paths in the form of road boundary,a joint control method of trajectory planning and tracking considering driving stability is proposed.By designing an approximate distance calculation method,the calculation efficiency of the relative relationship between the vehicle and the path is improved.Combining the safe driving envelope and the tire friction circle,a linear joint constraint of lateral and longitudinal driving stability is constructed to improve the driving stability.The simulation results demonstrate its good self-planning ability.In addition,the influence of the road adhesion coefficient is considered in the tracking control,which avoided the tire slip,and enables to make use of the road surface to exert the car performance.