Fast and Easy Systematic and Stochastic Odometry Calibration

Abstract

A method of odometry calibration is proposed and validated which is designed to be as convenient as possible. Convenience is enhanced by reducing both the amount of software to be written and the amount of measurements to be made to a minimum. The path dependent nature of odometry can be exploited to reduce the amount of ground truth information to as little as a single known point. Existing odometry and covariance estimation functions themselves would be used to extract first order parameter variation. Linearization would be performed about the approximate available trajectory rather than the unknown ground truth one. While multiple trajectories would be required to calibrate variance models, it would not be required that they be the same or even similar. The technique is general enough to apply to any form of odometry and it is general enough to be used for the extraction of unknown parameters of either systematic odometry models or stochastic error models. The derivation and experimental validation of the technique are presented.