Research On Key Issues Of The Steering Torque Feedback System

The steering torque feedback system has been studied for a very long time.As the development of the vehicle technologies,especially the vehicle intelligent technologies and steer-by-wire technologies,many new requirements have been proposed for the steering torque feedback system.As one of the connections between the driver and the vehicle,the role of the steering torque feedback system has been more and more important.In the research of the share control systems based on the steering feedback torque,how to design the steering feedback torque that meets the driver’s desired handling properties is one of the key problems of the system.In the research of the steer-by-wire systems,due to the abandon of the mechanical connections between the handwheel and the front steering wheel,more degrees of freedom have been proposed to the design and control of the steering wheel feedback torque.At the same time,because of the absence of the directional source of the road feel,the study of the modelling of the traditional steering feedback torque becomes more and more important.Besides,due to the limitation of the traffic regulations,road experiments of the intelligent vehicles are still difficult.That makes the driving simulator play an important role in the process of the development.As one of the key parts of the driving simulator,the steering torque feedback system has a crucial effect on the quality of the driving simulator.This thesis is supported by the Natural Science Foundation of China “Studies on Integrated Modelling and Control Methods for Intelligent and Electric Vehicles”,the National Key Research and Development Program of China “Studies on the Human-Machine Co-pilot control for Intelligent and Electric Vehicles”,and the University-Enterprise Cooperation Program “Driver-In-the-Loop System for the Estimation of the Advanced Driver Assistant System”.Based on the summary and analysis of the research status of the steering torque feedback system,four aspects are studied in this thesis,including the influence of the steering feedback on the driver handling properties,the simulation of the real vehicle’s steering feedback torque,the design and control of the desired steering feedback torque,and the platform design of the steering torque feedback system.Firstly,in the study of the influences of the steering feedback torque on the driver handling properties,the driver arm compensation mechanism and the driver’s desired handling properties are obtained.A driver arm mechanical model is established,and an identification algorithm for the driver arm mechanical parameters is proposed.By analyzing the variation of the driver arm mechanical parameters under the change of the steering feedback torque,the compensation mechanism is obtained.On the perspective of the whole driver-vehicle-environment(DVE)system,a quantification method for the driver’s desired handling performance is proposed,where the driver’s desired handling properties are represented by the DVE system transfer function under the driver’s inherent desired handling status.By analyzing the influences of the steering feedback torque and vehicle speed on the system transfer function,and the differences between the system transfer function under different driver’s handling,it is proved that the system transfer function can present the driver’s inherent desired handling properties,and can present the differences between drivers.Then,the quantification of the driver’s inherent desired handling properties has been realized.The system transfer function can be considered as the reference model of the steering feedback torque controller.Secondly,how to use the neural network(NN)to approximate a real vehicle’s steering feedback torque has been studied,and an easily applied steering feedback torque model based on neural network has been proposed.Firstly,based on real vehicle experiments,a great deal of experimental data are obtained,and multiple training datasets and prediction datasets are established.Then,based on these datasets,a steering torque feedback model based on Back Propagation NN is established.Based on this model,the relationship between the network framework and the network performance is studied,including the influence of the input variables on the fidelity of the model output,the influence of the number of the hidden neurons and hidden layers on the training efficiency and fidelity of the neural network model,the influence of the training dataset on the prediction performance of the NN model.After that,a basic framework of the neural network used for steering feedback torque modelling has been formed.Based on the framework,to further simulate the dynamic properties of the steering feedback torque,a steering feedback model based on Nonlinear Auto-Regressive e Xogenous input(NARX)neural network has been established.The time delay order of the input and feedback output has been studied.Finally,by comparing the prediction performance of the BP NN model and the NARX NN model,it is found that the NARX NN model used for steering torque feedback model is more effective,which provides basic steering feel for the study of the desired steering feedback torque.Thirdly,to improve the guiding function of the steering feedback torque,based on the researches on the influences of the steering feedback torque on the driver handling properties,a novel desired steering feedback torque controller is proposed.The desired DVE system transfer function is considered as the reference model of the controller.By controlling the steering wheel torque overlay,the handling performance of the whole DVE system is approximated to the reference model.Then,the driver handling load is reduced,and the performance of the whole system is improved.The controller is divided into two control loops,the inner and outer control loops.In the outer control loop,the desired steering feedback torque model is identified by an algorithm based on the least square method without need the detailed information of the driver.In the inner control loop,an adaptive neural network algorithm with a robust item is proposed to make the vehicle assembly behave like the desired steering feedback torque model.In the inner loop controller,the vehicle assembly is approximated by a Radial Basis Function(RBF)NN without need to know the detailed information of the vehicle assembly.The robust item can resist the system uncertainties and external disturbances.Therefore,the introduction of the desired driver handling performance and the realization of the desired performance is decoupled,and the detailed information of the driver and vehicle is not needed.Then,it can be concluded that the controller proposed in this paper can be easily used by different drivers and can be easily applied on different vehicles.Besides,the stability of the system has been analyzed.Finally,the effectiveness of the controller is proved by both the simulation and experiments,where the basic steering feel used in the experiments is provided by the steering feedback torque model based on the NARX NN.Fourthly,considering the stability influencing factors of the steering torque feedback system,a criteria of the system stability has been derived by the energy method.Based on the criteria,the influences of these factors are analyzed,including the time delay,the damping and stiffness of the mechanical part of the system,the stiffness of the steering feedback torque model,and driver input.Finally,the criteria and these conclusions of the influencing factors are verified by both simulation and experiments.This criteria can be considered as the design criteria used in the design of platform of the steering torque feedback system.The users can used this criteria calculating the range of the parameters of the system mechanical part combing with the range of the steering feedback torque rendered on the platform,which can guide the device selection and the design of the mechanical part.For an existing steering torque feedback system,the range of the steering feedback torque that can be simulated by the platform can also be calculated by the criteria combing the time delay and the parameters of the mechanical part.The torque range could make the user have a prediction of the existing steering torque feedback system.In addition,the experimental platforms needed by above researches have been provided in this thesis.A real vehicle experimental platform and a driving simulator around a steering torque feedback system have been constructed.Based on the real vehicle platform,the real vehicle experiments are designed,and based on the experimental data,the training dataset and prediction dataset for NNs are established.On the driving simulator,the experiments for the study of the driver arm compensation mechanism and the handling performance have been designed and conducted.Besides,the steering feedback torque controller,the system stability criteria,and the conclusions related to the influencing factors of the system stability have been verified on the driving simulator.