Multi-robot cooperative localization with optimally fused information of odometer and GPS

Abstract

We propose a data fusion technique for obtaining more precise estimation of positions of robots in common workspace, utilizing correlation between GPS on each robot. Because commercially inexpensive GPS shows several meters of deviation from true data, it is difficult for single GPS itself to contribute to mobile robot localization which needs several tens of centimeter accuracy for robot’s localization. With the experiments showing that GPS data of individual robot are correlated strongly as the distance between robots are close, we apply the concept of DGPS to the localization problems, and confirm that the proposed method provides improved localization accuracy. In addition, we derived the weight of sensor data to optimize fusing of odometer and GPS in multiple robot system. Applying optimal weight, we obtained improved localization performance through simulation.