Model-based path finding using adjacent area shape

This paper describes a robust and broadly applicable path-finding method that uses the range data of object surfaces adjacent to the path. The proposed method locates the path by using the shape defined by multiple sequences of less noisy parts of the range data. It is based on an originally devised algorithm for matching the shape of the object to that of the prototype model. This shape-matching algorithm can define the degree of similarity for curved surfaces three-dimensionally without assuming that the surface consists of parametric surfaces, such as planes, or conic surfaces. Segmentation of the surface is therefore not required. Moreover, this similarity can be defined even if the processing extent is not determined in advance. Consequently, the position of the paths used for robotic movement can be determined accurately from matching information by using less noisy parts of the range data, without putting restrictions on the shape of the object