A Practical Design towards Vision-based Self-localization System for Mobile Robot

In this paper, we focus on the practical design of on-the-fly self-localization system for mobile robot which operates in 3D environment. A concurrent general frame for this system is proposed, in which the information collected from odometer and unidirectional camera are integrated. We build the nonlinear models for these two physical sensors and design an efficient data structure for the CAD landmarks to guide the feature detection. Furthermore, the schemes of image feature extraction and the data association are presented. Experimental result is given finally.