A study of autonomous mobile system in outdoor environment. Part 5. Development of a self-positioning system with an omnidirectional vision system

Abstract

This paper presents a self-positioning system for a mobile robot. The proposed positioning system consists of an omnidirectional vision system (ODV) featuring two mirrors with a contrived curvature for obtaining a panoramic image with low aberration. An obtained panoramic image which occupies most of the visual field can be processed to recognize the topological position of three landmarks which are placed in the environment. A primary marker and two supplementary markers are used to indicate the correspondence between the real environment and the digital map. The relative self-position can be estimated from three directional angles toward landmarks. The absolute self-position can also be computed by comparing the landmark position on a map. Comparison between the theoretical positioning accuracy based on the Cramer-Rao lower bound and the experimental result in the field proves that both the ODV and proposed self-positioning algorithm including subpixel level image processing are valid and effective. An outdoor experiment which used a mobile robot shows that the proposed positioning system have high positioning accuracy even in the cluttered outdoor environment.