Robust range-based localization and motion planning under uncertainty using ultra-wideband radio

Abstract

The work presented in this thesis addresses two problems: accurately localizing a mobile robot using ultra-wideband (UWB) radio signals in GPS-denied environments; and planning robot trajectories that incorporate belief uncertainty using probabilistic state estimates. Addressing the former, we improve upon traditional approaches to range-based localization by capturing non-linear sensor dynamics using a Monte Carlo method for hidden bias estimation. For the latter, we overcome current limitations of scalable belief space planning by adapting the Probabilistic Roadmap algorithm to enable trajectory search in belief space for minimal uncertainty paths. We contribute a novel solution motivated by linear least-squares estimation and the Riccati equation that provides linear belief updates, allowing us to combine several prediction and measurement steps into one efficient update. This reduces the time required to compute a plan by over two orders of magnitude, leading to a tractable belief space planning method which we call the Belief Roadmap (BRM) algorithm.