Cooperative multi-robot localization based on distributed UKF

Abstract

In order to minimize the computational bottlenecks associated with centralized architectures for multi-robot localization, an algorithm based on distributed UKF for the localization of a team of robots is described. The scheme is that each of robots maintains a pose estimate using UKF that is local to every robot. When some of robots in the team can not receive the signal of GPS or does not set the GPS receiver, they can measure the relative distance and bearing towards other robots and use the relative observations to improve the precision of localization. Experiment based on simulation show the relative observation-based UKF is very effective in the heterogeneous cooperative localization.