Relative Localization in Multi-Robot Systems Based on Dead Reckoning and UWB Ranging

EasyChair Preprint no. 3705

Abstract

Combining dead reckoning and Ultra-Wideband (UWB) ranging information to achieve relative localization (RL) becomes prevalent in recent years. However, two main problems, i.e., pose initialization and distributed implementation, are rarely investigated in practical multi-robot systems. In this paper, a novel two-stage RL method is proposed to fill this gap, wherein an initialization strategy, using robot-to-robot measurements acquired at different vantage points during robot motion to determine initial pose, and a consensus-based distributed particle filter (DPF), fusing statistics from local robot and neighbors to realize RL, are designed. In our system, by computing the pairwise relative pose between all robots in a team, the initialization strategy can determine an estimate of the initial pose of all robots with respect to a common reference frame and the corresponding covariance. The consensus-based DPF allows determining the relative pose of all robots with significantly reduced computing costs. Experiments results on a team of differentially driven mobile robots show the effectiveness of the initialization strategy and highlight the low computation cost of the proposed approach.

Keyphrases: Distributed Particle Filter, Muti-robot, Pose Initialization, relative localization

@Booklet{EasyChair:3705,
  author = {Ming Li and Zhuang Chang and Zhen Zhong and Yan Gao},
  title = {Relative Localization in Multi-Robot Systems Based on Dead Reckoning and UWB Ranging},
  howpublished = {EasyChair Preprint no. 3705},

  year = {EasyChair, 2020}}