Task-oriented navigation algorithms for an outdoor environment with colored borders and obstacles

This paper presents task-oriented navigation algorithms used for an outdoor environment. The goals of the navigation are recognizing colored border lines on both sides of a path, avoiding obstacles on the path, and navigating the given path. To recognize the colored border lines with one camera, we apply a support vector data description method, which employs six color features extracted from two color models. To avoid collision with obstacles on the path, we fuse the data of the lines measured by a camera and the obstacles measured by a laser range finder. These algorithms were applied to autonomous navigation of about 100 m long curved track. We demonstrate that a four-wheel skid-steeringmobile robot successfully finishes the mission.