Pipelined Image and Data Processing for Robot Motion Planning and Navigation

This work presents a vision model structured as a pipelined image and data processing algorithm for robot motion planning and navigation, within semi-structured environments. The vision model emulates the way living creatures see their environment, and the way the acquired information is used to explore it. The proposed pipeline structure is divided into 3 abstract stages, each of which is further divided into more detailed blocks. Based on data acquired from captured images, a 2D map is formed, which can then be used for navigation. As new areas of the surroundings are explored, new data is obtained and the 2D map is updated accordingly. This paper establishes groundwork for future improvements of the image and data processing algorithm in a way that enables replacement of stages of the pipeline in an independent manner. Experimental results, obtained using a PIII class PC and a USB Webcam, are presented