Research On Path Tracking Control For Vision Based Intelligent Vehicle

As core part of the intelligent transportation systems, intelligent vehicle systems have become a hot research area for counties in the world under the increasingly high demand for safety and intelligent driving. Vision based intelligent vehicle are the most active and potential in the intelligent vehicle systems for their relatively simple structure and informative advantages. In this paper, monocular vision navigation based intelligent vehicle is studied, and the research work is carried out on the key technologies for lane detection and path tracking control.Firstly, under the principles of monocular vision for autonomous navigation, the intelligent vehicle system structure is designed especially in the key elements such as the steering, braking and drive system module. And due to key technologies in the lane detection for monocular vision and navigation, enhanced edge detection algorithm based on histogram threshold is proposed. Combined with the principle of Hough transform, lane detection is realized and the simulation results represents the method is efficient, simple and practical, easy to implement. Then calculate the deviation between the expectation driving trajectory and the road trajectory processed from the image acquired.Intelligent vehicle path tracking control based on vehicle dynamics is focused on study. Dynamic model takes full consideration of tire cornering stiffness and can observe lateral acceleration and vehicle yaw rate to evaluate path tracking performance. The path tracking model is established by vision preview theory, so that the vehicle can predict road ahead, and then control the vehicle ahead of time to achieve good tracking performance. The extended model combined vehicle dynamics and road preview is established.In order to achieve path tracking control, modern control theory is used to build quadratic functions for optional controller. Detailed research is carried out for the selection of the state weighting matrix to ensure the designed controller towards the closed-loop system to own good dynamic response. Step response, response with the initial error and robustness analysis is carried out to check performances of the designed optional controller. Results show that the controller has a good dynamic performance and robustness.Finally, SIMULINK model is built to analysis path tracking control for the vision based intelligent vehicle. Simulation results of the arc path tracking and lane changing path tracking represents the designed optional controller meets requirements for vehicle driving. Towards small radius of road tracking, curvature smooth algorithm is proposed. Results show that the method can reduce the lateral acceleration and yaw angular acceleration when path curvature changes largely in a short time, thus providing good driving comfort.