| As technology advances and tracffic safety issues become more serious,people have concentrated on autonomous vechicles,which is safer and more intelligent.However,the current technology in automatic driving is not mature enough to be directly applied to daily cars,whose control precision and safety assurance capability in complex driving situations shoud be further improved.This research focuses on path tracking control system,which is one of key technologies in automatic driving,and explores the related contents including parameters identification algorithm and path tracking algorithm.First,the whole vehicle system is decomcomposed into vehicle body model,tyre model and steering model.Considering the balance between model accurary and computing complexity,the four tyres vehicle body model suitable for identification and the bicycle vechcle body model suitable for control are established.To ensure the realtime ability of model parameters,a self-adaptive linear tyre model is developed.A simplified steering model based on steady state charactaristics is designed.Besides,under the consideration of the requirement of control system,the established vehicle models are integrated and utilized to derive the vehicle-road error dynamic model,which contains the road curvature information.Subsequently,to solve the problem of obtaining steering ratio,a kinematics based identification system including luenberger observer and recursive least squares method is designed,and the self-adaptive forgetting factor is added to enchance the system performance.Considering the real-time requirement of cornering stiffness in control system,based on four tyres vehicle body model,the identification systems of extended Kalman filter/unscented Kalman filter+Kalman filter are developed respectively.Meanwhile,the self-adaptive rule of varying parameters is adopted to improve the system accurary and robustness.Finally,the model predictive control is deduced and designed based on error dynamic model,and stronger stability of this algorithm is achieved by establishing dynamic constraints.Noted the computation efficiency of analytical solution,utilizeing the preview theory,the augmented linear quadratic regulator is designed to deal with the nonlinearity of error dynamic model.Meanwhile,in order to ensure the robustness of this control algorithm,a varying gain method is developed,which takes the constraints of mass center sideslip angle and tyre sideslip angle into account.Then,on the premise of simulated annealing algorithm and radial neural network,an optimal preview time selector is designed,which enchances the algorithm’s adaptability in varying velocities and road adhesion conditions.The theoretical analysis results of stability,steady-state response in the time domain and system response in the frequency domain verify the effectiveness of the augmented linear quadratic regulator with constraints.Eventually,a series of simulation tests guarantees the reliablilty and accurary of the designed model predictive control and the augmented linear quadratic regulator with constraints. |
