Research On The Identification Algorithm Of Driver Behavior Characteristics For "Car Adapts To Driver" X-by-wire Vehicle

Due to the restrictions on vehicle mechanical structure, traditional automotive control systems don’t have the great potential. With the rapid development of automotive networks and microprocessors, X-by-wire technology is increasingly used in the vehicle by many research institutions and automotive manufacturers. Since the hydraulic, mechanical and other devices or components are reduced, the quality of X-by-wire vehicle is reduced and the wire arrangement becomes convenient. With the flexible control algorithm and adjustable system parameters, the dynamics control of X-by-wire vehicle has larger development space than that of traditional automotive, which makes X-by-wire vehicle become domestic and foreign research focus. Introducing the drivers’characteristics to vehicle integrated control system, user-friendly design can be achieved, the form of "driver adapts to car" will be converted to "car adapts to driver". In addition, the applications of various electronic control systems improve the automotive electronics and intelligent level and play an increasingly important role for promoting active safety and driving comfort of the vehicle. In order to improve the drivers’acceptance toward driver assistance systems, drivers’characteristics should be considered to the design of control algorithms, finally achieving the driver adaptive and personalized driving, on the premise of ensuring vehicle safety and driving comfort. Therefore, the drivers’characteristics should be identified, whether it is to achieve the "car adapts to driver" X-by-wire Vehicle, or to improve the drivers’acceptance toward driver assistance systems.This paper relies on the proposal of the open fund project of the state key laboratory of automotive simulation and control——Integrated Control Method for a Full Drive-by-Wire Electric Vehicle Based on Driver’s Intention Recognition (project number:20120111), China postdoctoral science foundation(project number:2014M561289) and youth fund of national natural science foundation projects(project number:51305190). In order to establish the identification algorithm of driver behavior characteristics for chassis dynamics control system of X-by-wire vehicle, this paper has built a driving simulation, selected certain drivers to conduct experiments, extracted feature parameters from experimental data collected, classified driver behavior characteristics by use of K-means clustering method, and built identification models of driver behavior characteristics by use of data samples obtained after clustering based on Artificial Neural Network(ANN), and verified the accuracy abilities and the predictive abilities of the models by conducting driving simulator tests.In order to establish the identification models of driver behavior characteristics, this paper carried out the following tasks:(1) Built a driving simulatorA large number of experiments under different conditions should be conducted, in order to fully exploit driver behavior characteristics. Due to fewer adjustable parameters for real vehicle, as well as the designed conditions limited by environment factors, this paper built a small driving simulator to conduct experiments and to collect data samples which would be used to establish the identification model of driver behavior characteristics, based on the existing research in the research group. The paper presented the overall framework and the working principle of driving simulator, and introduced in detail the key components of driving simulator, including the cockpit, the console, the real-time dynamic simulation model, the real-time simulation system, the steering force feeling simulation system and sensor systems.(2) Classified driver behavior characteristicsThis paper summarized and analyzed the classification methods of driver behavior characteristics, selecting K-means clustering algorithm to classify driver behavior characteristics. The steering, braking and accelerating test conditions on driving simulator built are designed, and13drivers are selected to test and experiment data are simultaneously collected. Drivers’steering behavior, braking behavior and accelerating behavior are analyzed, and the feature parameters are selected and extracted from the experimental data by use of MATLAB. Then the feature parameters are clustered by use of K-means algorithm, thus the driver’s steering behavior characteristics, braking behavior characteristics and accelerating behavior characteristics are divided into the cautious type, the average type and the aggressive type, respectively. At the same time, the data samples of each type are obtained, preparing for building the identification models of driver behavior characteristics.(3) Established identification models of driver behavior characteristicsFrom the perspective that driver characteristics identification is to recognize the pattern of driver characteristics, this paper introduced several common methods of pattern recognition, and analyzed and compared their advantages and disadvantages, as well as their strengths and weaknesses and the scope for identifying driver behavior characteristics, before this paper finally selected ANN as the method of building identification model of driver behavior characteristics. This paper established identification models of driver behavior characteristics based on BP Neural Network by use of the data samples obtained during the driver characteristics classification. The variables that input and output to the network, the structure and the training process of BP neural network have been introduced in detail. And the driving simulator tests verified the accuracy ability and predictive ability of the models.