Curve path detection of unstructured roads for the outdoor robot navigation

Path planning is a crucial problem in the application of mobile robot and autonomous vehicle. Recently most of methods only achieve reliable results in some particular structured environments. In this paper, we describe a new navigation path detection algorithm in an unstructured environment. Natural frames are analyzed in RGB vector space to research the feasibility of curve path detection of unstructured roads. Perceptual color clustering and morphological image processing have been used as pre-processing to obtain the segmented path. The pixels of the segmented image are analyzed from right to left line-by-line to find the middle points of the path. In order to obtain the optimal navigation path, the least-squares curve fitting method has been used in our research. At last a new method of obtaining the navigation parameters is given. The experimental result shows that our algorithm has high speed and remarkable precision. The navigation parameters in our algorithm can control the robot movement well.