Ultra-Wideband Error Modeling for Improved Localization

Abstract

Ultra-wideband (UWB) is a modern range measurement technology which can provide high-speed, low-cost ranging, however UWB measurements can be difficult to model. In an effort to increase accuracy of localization using UWB, this thesis develops models to better understand the complex error patterns of UWB range measurements, specifically how separation distance and relative angle between modules affect error. These models are used to develop three error prediction and correction methods to improve localization: (1) range-based error correction, (2) angle-based error correction, and (3) fused range-angle error correction. While it was found that decreasing mean measurement error does not always decrease localization error, the lowest measurement error and lowest localization error both resulted from the fused error correction method. The fused error model combines the separation distance and relative angle models to predict and correct for range error, decreasing the mean measurement error by over 80%, the mean localization error by approximately 35% when using least squares estimation, and by approximately 56% when smoothing the trajectory with a Kalman Filter.