Learning-based Optimization And Extreme Drift Control For Lateral Stability Of Autonomous Vehicle

Autonomous driving technology is of great significance in reducing the occurrence of traffic accidents and improving driving safety.As the last link of autonomous driving technology,vehicle motion control directly affects the safety and reliability of autonomous driving.When the vehicle is driving in complex or extreme working conditions,it is easy for lateral instability to occur,leading to the potential for accidents.Therefore,ensuring is an important the lateral vehicles stability of autonomous conditions in different working topic.The lateral stability control of autonomous vehicles under different operating conditions is studied in this paper.Based on the analysis of vehicle working conditions,a vehicle lateral dynamics model considering tire nonlinearity was constructed under non-limit working conditions.Combining the advantages of various control algorithms,a vehicle lateral stability controller based on kernel approximate dynamic programming method and learning-based rolling optimization method was designed to ensure vehicle driving safety and improve control accuracy.The vehicle simulation software Carsim-Matlab co-simulation was used to verify the effectiveness of the control scheme under various working conditions.In extreme working conditions,considering the influence of tire saturation characteristics,a vehicle lateral stability controller based on vehicle drift characteristics was designed.The effectiveness of the controller was also verified by Carsim-Matlab co-simulation,which ensured the lateral stability of the vehicle in extreme working conditions.Firstly,in order to solve the problem that the vehicle tire force will enter the nonlinear region under complex operating conditions and the traditional stability control method depends on the accuracy of submodels,a lateral stability control method based on kernel approximate dynamic programming is proposed.A Pacejka tire model and a two-degree-of-freedom vehicle dynamics model were constructed,taking into account tire nonlinearity.Tire parameters are identified based on tire experimental data.A kernel dictionary is constructed using approximate linear correlation analysis method to collect vehicle data off-line.An executive-evaluation network is designed for vehicle stability control problems.The optimal weight updating iteration strategy,considering the time-varying forgetting factor of least squares algorithm,is studied.The effect of vehicle stability control is verified under different working conditions and compared with the control effect of Approximate Dynamic Programming(ADP),which provides a theoretical basis for the predictive control method based on learning.Secondly,in order to reduce computational cost and improve control accuracy,the kernel-based approximate dynamic programming algorithm is combined with the traditional Model Predictive Control(MPC)algorithm,and the rolling optimization characteristics of MPC are introduced into the finite-time domain approximate dynamic programming algorithm.The infinite-time domain problem is decomposed into finite-time domain problems and an iteration rule of time-varying forgetting factor based on rolling optimization is designed.A vehicle stability controller based on learning rolling optimization is designed.Simulation experiments are carried out under various working conditions,and the simulation results show that the learning-based rolling optimization algorithm has better control accuracy than the traditional ADP algorithm and MPC algorithm.Finally,aiming to address the problem that the traditional stability control strategy cannot guarantee the vehicle’s stability when the vehicle is in its ultimate working condition,a drift control strategy is proposed to ensure the lateral stability of the vehicle in this condition.A three-degree-of-freedom vehicle dynamics model considering tire saturation characteristics is constructed,and the steady-state drift equilibrium point and the vehicle drift characteristics are analyzed.The relationship between the vehicle’s drift attitude and drift path is studied,and a vehicle drift controller based on nonlinear model predictive control is designed.It is proved by co-simulation that the controller can realize the drift tracking of the desired trajectory at high speed,and ensure the lateral stability control of the vehicle in the limit condition.