Dead Reckoning of a Modular Omnidirectional Vehicle by Fusing Redundant Odometry and Gyro Information

Abstract

This paper proposes a dead reckoning of a modular omnidirectional vehicle. Wheel units assembled in the vehicle are modularized on their mechanisms and controllers, so that we can easily tailor the vehicle to individual transfer applicatoins. Each of the wheel modules estimates the vehicle position based on the information of the odometry and gyro. The wheel modules exchange their local estimates. The local estimates are fused in a decentralized manner, and then the wheel modules can precisely identify the vehicle position under the conditions of wheel slips and changes of wheel radius. The dead reckoning algorithm is formulated based on the extended Kalman filter, and some fusion rules are incorporated into the algorithm to improve the accuracy of the dead reckoning. Simulation and experimental results show that our dead reckoning can provide better positioning accuracy than the conventional dead reckoning.