Calibrating Away Inaccuracies in Ultra Wideband Range Measurements: A Maximum Likelihood Approach

Abstract

A calibration framework for an ultra-wideband localization system employing inertial measurement units is presented. No external motion capture system or other sensors are required for the calibration procedure. Given a covariance function for the error in the range measurements, a range measurement model based on a Gaussian process is obtained by maximizing the joint likelihood of angular rate, acceleration, and range measurements. The framework is experimentally evaluated, and it is shown how the resulting measurement model, integrated in a standard Kalman filter, can be used for real-time localization on a platform with limited computational resources. The calibration significantly improves the localization accuracy for randomly generated trajectories and different localization system setups.