Study On Driver Model For Adaptive Control Behavior Of Vehicle Direction

Main tasks in the development of modern motor vehicles are the increase of driving safety and comfort as well as the relief of the driver by driver assistance systems. Successful modeling and simulation of driver behavior is important for the current industrial thrust of computer-based vehicle development. Frequent traffic accidents and huge casualties and property losses make people come to realize the importance of the research of vehicle safety, particularly the importance of active safety systems. These accidents are the results of complex interactions between the driver, vehicle, and the environment. A study sponsored by NHTSA found that driver error was the major contributor in more than 90% of the crashes they examined. Furthermore, drivers could exhibit vastly different response under the same driving task. It is thus very important for automotive engineers and designers of vehicle active safety systems to understand, and simulate human driving behavior accurately. This makes the researchers start looking for an effective and accurate driver model, including all the characteristics of behavior of the drivers, in order to set up a complete and comprehensive driver - vehicle - road closed-loop model.First of all, concerning the learning process of driver, we study the directional control of driver behavior. Based on the driver-vehicle closed-loop system that has been set up, we design the control correction loop. People study the driving skills in the low-speed condition, since in this condition, the vehicle dynamics properties meets the simple linear relationships. As a result of being sensitive to the deviations between the current locus of the vehicle and road track, we introduce the PID controller with the error of these two signals as the input, simulating the directional control of driver behavior in low-speed condition. In order to simulate the capability of driver's self-learning and the process of accumulating the experience, we introduce the genetic algorithms in this article. Using the simulation results in the low-speed and slow turning driving conditions, we optimize the PID control design parameters, according to the principle of minimum driving error and physical burden. Then the process of optimizing the PID control parameters is simulated from the process that turns a real driver to be a qualified one through a series of driver training. Through combining the CarSim and Simulink software to simulate and test, it is confirmed that the directional control of driver model based on genetic optimization of off-line tuning in the low-speed conditions can describe the driver behavior in that condition perfectly.In addition, according to the process that turns people from a general driver to a skilled driver, we study the driver's directional control behavior. Based on the driver model that we have set up, we improve the control correction loop. In the high-speed condition, vehicles exhibit strong nonlinear characteristics. The driving experience from the low-speed conditions does not meet the high-speed condition. Skilled driver, however, in accordance with the non-linear response of vehicles which obtained from the original experience and the current vehicle trajectory deviations, could adjust their control behavior continuously, and thus control vehicle on the ideal track. From the adaptive mechanism of neural network and the strategy of skilled drivers managed to strong nonlinear of vehicle dynamics, we propose the rules that could adjust the above-mentioned PID control parameters online, and ultimately establish the driver model for adaptive control behavior of vehicle direction based on genetic optimization in offline tuning and neural network in online tuning. After simulation, the model can control the vehicle accurately in the expected trajectory in high-speed condition and simulate the understanding of skilled drivers to the non-linear characteristics of vehicles, making the model to be effectively applied to the driver - vehicle closed-loop system.Finally, we preliminarily study the driver preview behavior during driving. We define the process that drivers observe the path environment as the driver preview behavior, which is the most important source of information during driving. Therefore, the driver preview behavior should be introduced into the driver model in order to establish a precise manipulation that is in consistent with the real driver. In this paper, we set up the relationship between speed and road curvature and the preview time, in terms of the real driver in different curvature and speed condition of ground test, and apply this relationship in the driver model for control behavior of vehicle direction.In summary, we set up establish the driver model for adaptive PID control behavior of vehicle direction based on genetic optimization in offline tuning and neural network in online tuning. Through the ground test, we preliminarily explore the relationship between speed and road curvature and the preview time. This driver model can represent the real understand and experience of the real driver and accurately simulate the driving behavior of the real driver.Next, we will explore driver model that possesses the self-improvement and the memory function, as well as different driving styles.