Development of a car-like online navigation testbed

摘要

We present new realtime path planning and collision avoidance algorithms for an autonomous rover equipped with a laser range finder to be used as a platform for multi-agent navigation and control in unknown environments. For successful navigation, such tasks as localization, map-building, and collision avoidance should be handled at the vehicle level. The proposed architecture covers these aspects of robotic path- planning in a modular and robust manner, allowing quicker development of more sophisticated path-planners. Using a conventional SLAM algorithm, a feature map and the location of the vehicle is obtained. The information for orientation and distance of the obstacles ahead is available from a laser range finder. The proposed collision avoidance algorithm provides multiple paths to guide the vehicle through the environment. The system acts as a self-contained extendable platform for development and testing of high-level pathfinders.