Path Planning and Navigation in a Sparse World Space Environment

This paper investigates the mobile robot navigation problem in a situation where the information about the navigation world can be available in a form of a sparse data set. Such a problem arises when the information about the navigation environment, obtained usually in a raster format from a vision sensor, is subject to noise, continuity and connectivity distortions. In order to deal with the data sparseness, an approach based on spatial filtering, a derivative of a Gabor filter, is proposed. The results of the filtering procedure are used in a different way for path planning and for navigation tasks. The mobile robot path is defined in terms of a skeleton that retains the connectivity information of the shape of the admissible navigation area. The navigation algorithm applies gradient-based neuromorphic processing on the environment map obtained by using a procedure for adaptive thresholding of the filtered sparse data.