The Research On Multiple Model Adaptive Control Of Automatic Vehicle Driving

In recent years,the increase in income of residents has led to the prosperity of automobile industry and sales market.The car brings conveniences and problems to our life at the same time.Some problems have seriously threatened people's lives.It is a universal true that automatic driving will have a positive impact on traffic.Famous automakers have developed auto-driven cars for years and some Internet firms such as Google,Baidu,and LeEco have also entered this field.Some products have been tested on the roads.But the test conditions are good enough for driving.The actual driving environment is complex.It is necessary to solve the problem of driving under complex road conditions for self-driving cars.The following aspects have been studied in this article:Firstly,the vehicle dynamics,the driving force and the resistance are studied.The paper analyzes some factors of certainty and uncertainty and considers an acceleration resistance.According to the relationship between output torque and the driving effect,the control model of vehicle movement with maximum torque is developed.Secondly,the non-linear relationship between adhesion coefficient and slip ratio is discussed.A sliding mode controller is developed to track and control the optimal slip ratio based on the adhesion coefficient.An example is given to show the good effect of tracking the optimal slip ratio even when the adhesion coefficient has a vast change.Thirdly,the multiple model adaptive control with second level adaption,a new concept in adaptive control using multiple models,is used for the control of automatic driving.The new multiple model method is based on Lyapunov stability.Compared with the traditional adaptive control method,the number of models is significantly smaller.By setting the initial values of identification models appropriately,the plant parameter vector will lie in the convex hull of those initial value set.The new method provides an estimate which depends upon the outputs of all the models and can be viewed as a convex combination of estimates.So all the identification information will be fully utilized,and the problems of discontinuity and jittering which are caused by switching between models would not happen in this method.The S-function in Simulink is utilized to accomplish the modeling of controlled object and the multiple models adaptive control algorithm with second level adaption.The validity of the method is verified through simulation.The simulation results show a faster convergence and better error control performances,Finally,the professional vehicle dynamics simulation software,CarSim,is used to perform a vehicle adaptive control simulation through configuring the variations of road and adhesion conditions.The results are consistent with the aforementioned theoretical simulation models.