Robust Loop Closure Selection for Multi-robot Mapping under Perceptual Aliasing

Abstract

Robust loop closure detection is crucial in simultaneous localization and mapping (SLAM) as failure to correctly choose the loop closures can severely distort the map. This applies not only to single-robot cases but also for multi-robot systems. When initial relative poses between robots are unknown, inter-robot loop closures must be found in order to merge their local maps and produce a global map. This becomes an even more challenging issue in the presence of perceptual aliasing that arises from repetitive features and patterns in highly structured settings such as indoor environments. In this paper, we propose a robust inter-robot loop closure selection method that finds the inlier loop closure set that maximizes the consistencies between measurements. We show that this can be formulated and solved as a maximum weight clique problem in graph theory. To demonstrate the performance of the algorithm, simulation is performed using synthetic data and the result is compared to a recent method dealing with the same problem.