Application of Segmented 2D Probabilistic Occupancy Maps for Mobile Robot Sensing and Navigation

The concepts of occupancy grids and probabilistic maps were introduced at the end of the eighties. Since then, research work focused mainly on the definition of the representation, data fusion and generation of occupancy models. Few consideration has been given to processing occupancy maps as textured images in order to extract meaningful information required for robot navigation and control of interactions with the environment. This paper investigates the application of segmentation techniques on probabilistic occupancy maps represented as textured images. Enhancements are proposed to a uniformity estimation technique based on local binary pattern and contrast (LBP/C) to achieve robust segmentation of occupancy maps that typically result from range sensors with limited resolution. The accuracy of the segmented 2D occupancy maps is demonstrated experimentally through an application on mobile robot navigation with collision avoidance