Research On Adaptive Model Predictive Lateral Control Method For Unmanned Ground Vehicles

Path tracking control is one of the key technologies of unmanned driving technologies,and it is essential to ensure path tracking performance and vehicle driving safety.Most of the existing controllers are designed based on fixed parameters,so they have insufficient adaptability to different working conditions.Especially in the path tracking scenes of complex road conditions at high-speed in an off-road environment,it is difficult for the existing controllers to meet the requirements of tracking accuracy and stability at the same time.Aimed at the above problems,the dissertation proposed an adaptive model predictive control method.The dissertation established a parameterized reference path model based on the Frenet coordinate system,introduced the "magic formula" to analyze and model tire dynamics,then analyzed the lateral dynamics of the vehicle,and established the vehicle dynamics model with tracking errors as the state variables.Then the dissertation briefly introduced the basis of the model predictive control algorithm,and carried out the design of the linear time-varying model predictive controller,and then based on the above work,the following improvement studies were made on the controller design:Firstly,aimied at the problem of insufficient adaptability of the existing model predictive controller to different working conditions,the dissertation analyzed the influences of key parameters such as time domains and constraints on the controller,and an model predictive controller with optimized parameters is designed.The dissertation established a parameters matching table from the optimized predictive time domain and control time domain parameters measured in the experiments,derived and analyzed the rotation angle constraints within the stability range.In actual application,the controller obtains the vehicle speed and road information through the real-time status of the unmanned vehicle,and selects the optimal predictive time domain and control time domain parameters,adjusts the rotation angle constraints,realizes a parameters adaptive mechanism to improve the tracking performance.Secondly,aimied at the typical working conditions of high-speed path tracking in off-road environment,considered the impacts of unmodeled environmental factors such as terrain and soil conditions and model deviations on the control effect,the dissertation designed a selflearning feedforward compensation controller,which forms a feed-forward-feedback controller structure with the parameters adaptive model predictive controller.In the controller structure,the learning coefficients of the feedforward controller are updated online according to the realtime status errors of the unmanned vehicle.The controller structure can quickly reduce the tracking errors when the errors are large,which can improve the tracking accuracy while ensuring stability.Then,the dissertation built a joint simulation platform,set up the typical Double Lane Change path,carried out the simulation verifications of different speeds and different slip conditions under typical paths for the improved controller respectively,the simulation results show that the designed controller can track the reference path relatively stably under different working conditions.Finally,the dissertation carried out the simulation comparison experiments for the designed controller and the traditional model predictive controller of the typical Double Lane Change path tracking under different slip conditions.And,the experimental verifications under the typical path conditions of the off-road scenes are carried out based on the actual vehicle experimental platform.And the results show that compared with the traditional model predictive controller,the tracking accuracy and stability of the designed controller are greatly improved.