UWB-Based Self-Localization Strategies: A Novel ICP-Based Method and a Comparative Assessment for Noisy-Ranges-Prone Environments

Abstract

Ultra-Wide-Band (UWB) positioning systems are now a real option to estimate the position of generic agents (e.g., robots) within indoor/GPS-denied environments. However, these environments can comprise metallic structures or other elements which can negatively affect the signal transmission and hence the accuracy of UWB-based position estimations. Regarding this fact, this paper proposes a novel method based on point-to-sphere ICP (Iterative Closest Point) to determine the 3D position of a UWB tag. In order to improve the results in noise-prone environments, our method first selects the anchors' subset which provides the position estimate with least uncertainty (i.e., largest agreement) in our approach. Furthermore, we propose a previous stage to filter the anchor-tag distances used as input of the ICP stage. We also consider the addition of a final step based on non-linear Kalman Filtering to improve the position estimates. Performance results for several configurations of our approach are reported in the experimental results section, including a comparison with the performance of other position-estimation algorithms based on trilateration. The experimental evaluation under laboratory conditions and inside the cargo hold of a vessel (i.e., a noise-prone scenario) proves the good performance of the ICP-based algorithm, as well as the effects induced by the prior and posterior filtering stages.