Longitudinal And Lateral Coupling Control Of Distributed Drive Electric Vehicles Based On Vehicle Parameters Estimation

As global environmental and energy problems become more and more serious,the electrical vehicle has become the trend of automobile technology.In recent years,electric vehicles have become the focus of scholars’ attention.The distributed drive electric vehicle has accurate and rapid driving/braking response due to the hub motor installed in four wheels,and is quite suitable for implementing various control methods to improve vehicle dynamical property.Therefore,the realization of vehicle yaw stability and high-precision trajectory tracking under different speeds and roads deserves intensive study.Obtaining accurate vehicle state parameters and road surface information is the premise of vehicle stability control.In this dissertation,a seven degree of freedom vehicle dynamic model is established for distributed drive electric vehicles.The longitudinal speed,yaw rate and side slip angle are estimated based on the improved RBF with K-means neural network.According to PCA analysis method of dimensionality reduction of high-dimensional data in machine learning,the main feature parameters are extracted for road adhesion coefficient estimator.It is shown that the estimation accuracy of both the side slip angle and the road adhesion coefficient is improved after filtering.The research results show that,compared with DRBF method,the estimation accuracy of side slip angle of DRBF-EKF method is improved by 68%,and the estimation accuracy of road adhesion coefficient is improved by 79%.For distributed drive electric vehicles,the conventional on-board sensor signals such as front wheel angle,longitudinal vehicle speed,longitudinal acceleration,lateral acceleration and wheel rotation are used as inputs,and a Levenberg Marquarelt Multi Module Self-Organizing Feedforward Neural Network(LM-MMSOFNN)based on the improved LM algorithm is proposed,which can estimate the road adhesion coefficients of the left and right wheels simultaneously.In order to improve the adaptiveness of the algorithm to the different road surface,the increase and decrease of self-organizing neurons is used to provide the necessary parameters for the vehicle stability control and trajectory tracking.Simulation and real vehicle test results show that both the LM-MMSOFNN method outperforms the Kmeans method,with an average absolute error of 0.0095 for the left side and 0.0138 for the right side.When the vehicle is turning,there is a large deviation between longitudinal and lateral forces of front and rear wheels.In order to obtain more accurate tire force information,a multi module self-organizing neural network tire force estimation method based on improved LM(LM-MMSOFNN)algorithm is proposed which can estimate longitudinal,lateral and vertical tire force simultaneously.The road recognition accuracy is improved by classifying road types based on self-organizing feedforward neural networks.Extended Kalman Filter(EKF)and moving average method are used to denoise and preprocess the measured signal and the computational complexity can be reduced by using the improved Levenberg Marquardt(LM)learning algorithm.The simulation and real vehicle test results show that the estimation errors with normalized root mean square error(NRMS),Root Mean Square Error(RMSE)and Mean Absolute Error(MAE)as evaluation indexes are significantly reduced by 12.3%,25.8% and 33.2%,respectively.A multivariable controller is designed for the coupling control of yaw stability and speed tracking of distributed drive electric vehicles.A fuzzy sliding mode control(Fuzzy-SMC)structure control is used to estimate the gain of the system to eliminate uncertainties and chattering.Hierarchical control method is adopted,the upper layer calculates the additional yaw moment with vehicle longitudinal speed tracking,and the lower layer is torque distribution layer.Considering motor characteristics and longitudinal force requirements,this control problem is transformed into a multiobjective optimization problem.In order to improve lateral tracking bias,a NALMPC trajectory tracking control method is proposed.Considering speed tracking,linear parameterization theory is introduced to discrete the vehicle dynamics model,and the step size of the prediction time domain is changed adaptively by vehicle longitudinal speed tracking to achieve the accuracy of trajectory tracking under different vehicle speeds and road adhesion coefficients.It is shown that the proposed algorithm is able to maintain vehicle stability and achieve accurate control of vehicle trajectory tracking under complex conditions of variable vehicle speed and variable adhesion coefficient.Aiming at vehicle state parameters estimation,road adhesion coefficient identification and tire force estimation of distributed driving vehicles,a high precision algorithm is proposed to realize the adaptive trajectory tracking control of vehicle under variable vehicle running speed variable and variable road adhesion coefficient.The accuracy of trajectory tracking control is guaranteed without increasing the burden of online calculation,and the proposed methods provide theoretical and technical support for safe driving of intelligent vehicles.