Visual navigation along reference lines and collision avoidance for autonomous vehicles

This paper describes the algorithms for generation of paths and lateral control of an unmanned vehicle with a stereo vision system for autonomous navigation and collision avoidance. One camera of the stereo vision system detects reference lines along which the vehicle is guided, and the stereo vision system detects obstacles. The lateral control consists of steps of the reference line detection, obstacle detection, path generation, and calculation of lateral control. The coefficient of the cubic curve that approximates the series of the sampled points provides the lateral control. Simulation studies and experiments were conducted to show the feasibility of the algorithms.