Self-Calibration in Differential Drive Dynamics/Kinematics Model

Author(s): Martins Ekmanis

Abstract: Localization and path planning is an essential elements of mobile robots but it relies on model availability. It is difficult to obtain robot dynamics/kinematics model due to lot of constants and variables. This paper proposes a highly adaptable and universal method for obtaining differential drive robot model by self-calibration. The obtained model is based on linear time invariant assumption and accepts sensor-less prediction as well as hybrid mode, where encoder’s data are filtered against the model. Theoretical background and experimental proof of model quality is included in this paper. The algorithms are implemented in software code and experimentally verified on physical robot. The experimental results confirm models accuracy by means of model based track following tests

Keywords: dynamics model; self-calibration; localization; robotics

Pages: 54-59

Download PDF