Mobile Robot Localization Based on Topological Model with Tiered Features

Perceiving the position of the mobile robot in an environment is an important element for an autonomous or semi-autonomous robot. In this paper we present a novel localization paradigm using topological map with tiered features, which is robust to rotation, illumination variations and scale changes. The topological model is extracted from natural landmarks and the normalized Zernike moments are calculated as main landmark features. By color matching and Zernike moments matching based on a probabilistic voting method between input scene and image models, natural landmarks can be recognized. The result is then fused with sonar data. The technique is especially adapted to indoor environment. Experimental results validate the favorable performance of this approach