Compact Pose-Graph SLAM Framework Based on Algebra with Embedded Uncertainty

Abstract

Simultaneous localization and mapping (SLAM) is the process of creating a map of a previously unknown environment while keeping track of the position of the mapping agent all the time as well. First a SLAM framework is being presented that provides a compact and therefore resource friendly method for storing maps, offering the possibility to implement single agent SLAM. The underlying method is a novel algebra, that represents 2D-vectors and points as three-element structures, having the uncertainty embedded as third parameter. Mapping is realized by detecting landmarks and storing their position, while the map is stored as a graph, with relatively tiny memory footprint. The paper first describes the framework and then provides simulation results using a differential-driven agent model.