| The safe navigation of vehicles in any environment requires the availability of an obstacle detection system for collision avoidance. Applications like roads, mining and space autonomous navigation demand the presence of such a system. The methods of obstacle detection are based on different types of sensors: camera, sonar, laser range finder and radar. Nowadays, laser range finder is the most studied sensor because of its direct access to scenes geometrical information.; In this work, we used a laser range finder for its range and angular accuracy and data acquisition speed qualities. We present a fast algorithm for obstacle detection on the path of an autonomous vehicle. Knowing the trajectory to be executed by the vehicle, its velocity and its dimensions, the obstacle detection module defines first a collision zone around this trajectory, in order to limit the search of the presence of obstacles. An emergency zone is also defined to emit a stopping command whenever an obstacle is found. Furthermore, obstacle detection is completed with occlusions detection thus enabling it to improve its efficiency. An occlusion occurs when a section of the collision zone is hidden from the sensor by an element of the scene not on its trajectory. After an occlusion is detected, the obstacle detection module sends a command to reduce the vehicle's speed to avoid colliding with an obstacle potentially present in the occluded region.; The proposed system was developed for the autonomous navigation on natural terrains having positive and negative obstacles such as holes. A laboratory and exterior experiment permitted the validation of this system. |
