Embedding local metrical map patches in a globally consistent topological map

Abstract

This article considers some practical and theoretical issues in the trade-off between globally consistent navigation and local precision manoeuvring. Precise local metrical maps are the common base for docking, manipulation, or other exact trajectory planning and control tasks. Yet these models are not scaling fine in the total geometrical size, when handling real world sensory data, and drifts. Nevertheless there is a need for a globally consistent spatial model for long term navigation. The presented work proposes a method of embedding local metric area patches in a topologically consistent global structure suitable for qualitative, and robust navigation. A global positioning information is not required at any stage, which limits the global precision of the spatial model, but on the other hand recommends it for environments where this information is not available. Results from physical experiments with autonomous robots are presented to demonstrate the robustness and practicality of the approach.