Research On The Driver Steering Characteristic Parameters Identification And Shared Control Algorithm For Path Tracking

With the rapid development of multi-sensor information fusion technology,artificial intelligence,5G communication technology and the continuous improvement of road infrastructure,intelligent vehicle based on electrification,intelligence and networking,has become a new trend in the development of automobile industry in order to solve driving safety,energy exhaustion and other problems.Therefore,the major automobile companies and technology companies have been devoted to self-driving technology.However,it is hard to complete self-driving technology in a short term.Semi-autonomous vehicles will be dominant in the future.In the semi-autonomous driving technology,the human-machine shared steering control has been considered to be the most potential assistance system to improve driving comfort,driving safety and vehicle handling stability.Human-machine shared steering control is also one of the research hotspots among advanced driving assistance system(ADAS)technologies.The key of this technology lies in machine’s understanding of the driver’s steering behavior.The driver’s steering behavior model can parameterize the steering behavior in order to design the shared controller.Therefore,it is of great significance to study the driver’s steering behavior model for human-machine shared steering control.This paper mainly studies the driver’s steering behavior and the robust control algorithm of human-machine shared steering path following.The main contributions in this paper are as follows:(1)A classification method for driver’s steering behavior with large curvature path following is proposed.Based on the six degree of freedom driving simulator platform,a large curvature road model and test scheme is established and designed.Some drivers of appropriate age are recruited to conduct steering test on this platform.The experimental data is used to extract the initial features of driver’s steering behavior.The correlation between these features is analyzed based on the initial features,and then the new dimensionality is obtained by principal component analysis(PCA).These new features are regarded as the inputs of k-means algorithm in order to cluster the driver’s steering behavior,and finally three types of driver’s steering behavior are obtained.(2)A driver’s steering behavior model is presented and driver’s steering characteristic parameters are identified.This paper combines mechanism of the driver’s steering behavior and assumption of two-point preview strategy in order to establish the driver’s steering behavior model.This model includes brain pure delay,arm muscle delay,steering preview gain,arm muscle stiffness and other driver’s steering characteristic parameters.Then,these characteristic parameters are identified by radial basis function(RBF)neural network.Based on identification results,the identification ability of this neural network is verified and analyzed.(3)Based on classification result of driver’s steering behavior,the boundary of driver’s steering characteristic parameters is divided.In addition,it is found that the characteristic parameters affecting the driver’s steering behavior are mainly the driver preview steering gain and feedback gain;no matter which driver’s steering category is,the driver’s response time and muscle stiffness parameters are basically the same,and there is almost no parameter perturbation with time.Therefore,in the process of controller design,only preview steering gain and feedback gain are considered to reflect the uncertainty of driver’s steering behavior,reducing the complexity of the controller(4)A fuzzy shared robust controller,considering the uncertainty of driver’s characteristic parameters and longitudinal speed,is designed.Firstly,the driver-vehicle-road model is fuzzed,and the output feedback is chosen as the compensation method of the controller.In addition,vehicle handling stability,path tracking performance and driving comfort are taken as the control objectives of the controller.These control objectives are optimized by classification result and parameter boundary.Finally,a personalized shared steering controller is designed.The shared controller is solved by the linear matrix inequality toolbox in MATLAB.Besides,the effectiveness of the controller is verified by the joint simulation of MATLAB/Simulink-Car Sim.