Implementation of autonomous vehicles with the Hough transform and fuzzy control

Abstract

An automatic driving system for a golf car is developed and tested in the campus. The system utilizes a single vision sensor CCD and adopts the road-model-matching method to obtain the lane marks by applying the Hough transform method. Compared with the inverse perspective mapping (IPM) method of which is affected greatly by the position, yaw and tilt angle of the camera [1] with heavy computation, the present method directly detects the lane marks without including information of the camera. Besides, the present method can be applied to detect curved lane marks simply by using a 2(rd)-order road model. Moreover, in the control for the nonholonomic vehicle, the commonly used PID control is not suitable for the automatic driving system. This paper proposes two different fuzzy logic controllers (FLCs) with nine and fifteen rules, respectively, to achieve the autonomous driving performance satisfactorily.