Automatic Steering Control For Autonomous Vehicles

In face of the increasingly severe traffic safety problem, people are attaching more and more importance to the study of intelligence vehicle. The present active safety technologies, such as Anti-lock Brake System(ABS) and Electronic Stability Program(ESP),though can improve driving safety considerably, but have not changed the traditional driving mode, thus can’t essentially stop the driver from making dangerous operations.Taking these factors into consideration, this paper puts focus on the study of the unmanned ground vehicle, laying stress on its automatic steering control.The current research status and related issues in the traditional automatic steering control methods are studied first. To improve the situation, a two-dimension vehicle road model which considers both the size and shape of the vehicle and road, is developed, and based on it, a new path tracking mode that names regional path tracking is proposed.In order to solve the proposed path tracking problem, the model predictive control method is adopted for designing the automatic steering controller. It is achieved by controlling the lateral motion of the unmanned ground vehicle. Thus the model that characterizes the plane kinematics and lateral dynamics of the vehicle is developed too.The road information obtained from the vision system is a series of position points, it need to be processed before it can be used as reference inputs to the automatic steering controller. For this, this paper introduces the quadratic search based three times Lagrange interpolation formula, and then develops an automatic steering control system based on the preview-follow theory. Simulations are conducted to test the proposed method,simulation results indicate that the model predict control based steering control system has good performance in dealing with the regional path tracking problem.Considering the model accuracy and control parameters may affect controller’s performance, this paper modifies the proposed control algorithm in three main aspects: First,establishes an error compensation control model based on the idea of disturbance estimation; Then, based on the new error compensation control model, equation of new steering control algorithm is derived; Finally, strategies of variable sample time and variable predictive time are proposed to further improve the performance of the model predictive control based steering controller. The superiority of the modified controller is demonstrated by simulation contrast verification.Real vehicle tests on experimental car HQ430 is conducted to test the practicability of the model predictive control based automatic steering control system. The main sensor used in the experimental test is an Inertial and GPS Navigation System RT3002. The experimental algorithms are programmed using C language, where the discretization of the model is simplified by Taylor expansion and the optimization is achieved via the differential evolution algorithm. Experimental results indicate that the model predictive control based steering controller satisfies the real time request of vehicle control and has a satisfactory control effect.