Intelligent neuro-fuzzy dynamic path following for car-like vehicle

Wheeled vehicles path planning and following are generally considered different phases. In this paper, an intelligent control system for dynamic milestone path following of four wheels car-like vehicles is presented. It is dynamic since, in order to be robust with respect to noise, using a milestone-path, the path between two consecutive points is planned only when the vehicle reaches the assigned milestone and the vehicle speed is computed during the following. It uses the path shape data, obtained in the planning phase, for the following phase. These data are used by neuro-fuzzy controllers trained for the path shape characteristics in order to calculate optimal wheels speed and front wheels orientation angle. These controller inputs are computed using both vehicle state as feedback and path shape as reference. Tests with Virtual Vehicle Manager control and simulation environment demonstrate the robustness of the real-time control system even in presence of noise and without exact knowledge of all the vehicle models and parameters.