Research On Path Tracking Control Of Intelligent Vehicle Based On Adaptive Two-point Preview

The popularity of automobiles not only makes people enjoy convenient travel,but also brings various problems such as congestion and collision accidents.Intelligent vehicles provide the best feasible solutions to these problems,among which,automatic driving technology is one of the important technologies of intelligent vehicles,and path tracking control technology is the key of automatic driving technology.Therefore,a good path tracking control technology can help solve the above problems.At present,the method of path tracking by preview is widely used by a large number of research institutions and technology enterprises.However,the preview distance has a decisive influence on the control effect of this method.How to determine the appropriate preview distance to make the controller have a good control effect is a problem worthy of further study.In this paper,aiming at the preview control method,a path tracking control method based on two-point adaptive preview is proposed.The main contents are as follows:(1)In order to meet the design requirements of the control system,a seven-degreeof-freedom model considering lateral,longitudinal,yaw and four-wheel rotation is built in Simulink.The seven-degree-of-freedom model can better reflect the key state of vehicle in the process of path tracking,and the complexity of the building is not high.Compared with the dynamic model provided by Carsim,the correctness of the sevendegree-of-freedom model is verified,and the theoretical thickness of the model is established.Then,according to the control objective and consideration of real-time,the seven-degree-of-freedom model is simplified to a two-degree-of-freedom dynamic model,which only considers yaw and lateral motion.The correctness of the two-degree-offreedom model is verified by Simulink and Carsim too,which provides a model basis for the design of controller.(2)A path tracking controller with hierarchical structure is established by using preview method.The upper layer uses two-point preview model and fuzzy reasoning to transform the lateral deviation and heading deviation at preview point into comprehensive deviation,so as to better describe the position state of the vehicle relative to the target path.Based on the assumption that the vehicle makes uniform circular motion in the preview period and the comprehensive deviation,the desired yaw rate can be obtained.The lower layer uses RBF(Radial basis function)neural network to adjust parameters of sliding mode controller in real time,and constructing neural network sliding mode controller to suppress chattering.The controller is used to track the target path.(3)The influence of preview distance on controller performance is analyzed,and the multi-objective optimization model of preview distance is studied.Firstly,the characteristics of particle swarm optimization and genetic algorithm are analyzed,and a hybrid particle swarm optimization algorithm is constructed.Then,combining the hierarchical path tracking controller and hybrid particle swarm optimization algorithm,a multi-objective optimization model of preview distance is constructed.The optimization model is used to iteratively calculate the optimal preview distance under different vehicle speeds and road radius.Finally,a dataset containing the optimal preview distance under different vehicle speeds and road radii is established,and the real-time optimization of preview distance is realized by querying the dataset.(4)A path tracking control method based on two-point adaptive preview is constructed by using the dataset containing the optimal preview distance and the path tracking controller with hierarchical structure.Using Carsim and Simulink to build simulation conditions.The algorithm proposed in this paper is compared with the traditional algorithm of adjusting preview distance according to the vehicle speed.The results show that,in the established simulation conditions,the proposed algorithm has improved driving stability and tracking accuracy compared with the traditional algorithm.When the vehicle speed is 20km/h,40km/h and 50km/h,the maximum lateral deviation decreases by 7.2%,17.9% and 22.3% respectively.