Abstract: Intelligent vehicles have become a research hotspot in academia and industry, mainly because the driving tasks of intelligent vehicles can be performed by the auto drive system, which greatly reduces the driving burden of human drivers and improves driving efficiency and safety. A large number of traffic jams and casualties are caused by the factor of “people”, and automatic driving can eliminate the hidden danger of “people” from the “people-vehicle-road” system, thereby greatly enhancing the safety of road traffic. Relevant research shows that the difference between an autonomous vehicle and a skilled driver is that the former gives people less comfort in steering and other operations. As one of the key operations of vehicles in the driving process, lane changing puts forward higher requirements for the transverse and longitudinal control of vehicles. In order to improve the tracking accuracy and occupant comfort of an intelligent vehicle in autonomous lane changing, this paper proposes a control method of intelligent vehicle lane changing by imitating human beings, which combines the extension control with TSK (Takagi Sugeno Kang) fuzzy control. First of all, five driving school coaches with rich driving experience are recruited as representatives of skilled drivers, the lane change track data of skilled drivers are collected by a combination of real vehicles and driving simulators, and the test track is analyzed to study the impact of different factors on lane change path. The behavior of intelligent vehicle lane changing has strong nonlinear characteristics. Based on generalized regression neural network (GRNN), a path planning model for human like lane changing is designed. Secondly, the relationship between preview position deviation and the yaw angle is derived so as to establish the “vehicle road” system model by analyzing the preview characteristics of skilled drivers and combining the preview deviation, road environment and vehicle state. In order to improve the tracking accuracy and occupant comfort of intelligent vehicles under different lane changing conditions, a humanoid trajectory tracking controller based on extension theory is designed. The controller can be divided into upper and lower layers. The upper layer controller mainly extracts the tracking characteristics of intelligent vehicles and divides the classical domain, extension domain and non domain. In order to meet the control requirements of intelligent vehicles in different control domains, the lower controller adopts PID feedback control based on road curvature and emergency braking respectively, which solves the limitations of a single control method and realizes accurate control in the global range. Because the extension control will produce control quantity jitter at the switching point, and the local stability will be affected, an optimization method based on TSK fuzzy theory is proposed to further improve the global stability and tracking accuracy of the extension control. By analyzing the results of joint simulation, it can be concluded that the extension control method based on TSK fuzzy theory proposed in this paper has higher control accuracy and stability.