Study On Classification And Identification Method Of The Driver’s Steering Characteristics

With widespread applications of intelligent control algorithm and the emergence of thefour-wheel independently driven in-wheel motors electric vehicle, it is possible to make theintelligent electric vehicle come true. And the intelligent electric vehicle has led gradually thetendency of the development of vehicle. Among all forms of the intelligent electric vehicle,the four-wheel independently driven in-wheel motors electric vehicle is more prospective.This thesis is supported by National Natural Science Foundation of China (GrantNo:51305190) which is titled with Study on Adaptive Steering for Intelligent Four-wheelIndependently Driven In-wheel Motors Electric Vehicle. In order to create conditions offurther researches on controlling in-wheel motors electric vehicle’s four-wheel drive torque tomake its steering consistent with driver maneuvers, and improve vehicle’s steering comfort,the classification and online identification of driver steering characteristics category aremainly studied in this thesis.According to the need of researches, a kind of driving simulator experimental platformwhich can acquire data, process data and display images has been built by applying CarSimRT called real time simulation software and dSPACE named real time simulation system. Thedynamics of steering process has been analyzed to determine measurements. Based on thisexperimental platform, a steering experiment has been designed. And then the experimentaldata corresponding to measurements have been acquired in this platform and processedartificially. Thus the steering data related with steering parameters have been achieved. Onthe basis of the steering data, features of driver steering characteristics have been investigatedreasonably by correlation analysis and values of features are also acquired by processing thesteering data. Three kinds of driver category, the cautious, the average and the reckless, havebeen ascertained by referring to research results from America and considering currentconditions. Characteristics of categories have been analyzed. The steering data are clusteredinto three clusters by fuzzy C-means clustering algorithm. Based on the aforementionedcharacteristics of categories, every diver category can have a one-to-one correspondence toeach clustering center of clusters and the sample data with category labels can be acquired.Training samples with proper noisy are extracted from sample data to train the identificationmodel of driver category built by BP neural network. Test samples are extracted from the restof sample data to test the performance of the model. The results show that all accuracy of thismodel to three categories are more than80%.Based on driving simulator experimental platform, online data processing program iswritten by applying Matlab/Simulink and RTI software. And the identification model isembedded into it. Several driver steering characteristics categories are tested and verified.The results turn out that it is possible to realize the online identification of driver steeringcharacteristics categories with sound real-time performance by using the method mentioned.