Multi-resolution and Particle Filter Based Global Cooperated Localization Method
YIN Peng,HE YuQing,HAN JianDa,XU WeiLiang
DOI: https://doi.org/10.1360/n092017-00422
2018-01-01
Abstract:Cooperation & coordination among heterogeneous robots (e.g., UAV & UGV, UAV & USV) have obtained extensive attentions due to its greatly enhanced perception ability, especially when facing complicated and unstructured outdoor environments. For multi-type robots cooperation task in the GPS-denied environment, cooperate localization is one of the most important task with great challenges which combine the aspects of data collection of different type sensors, adaption to the environment feature and perspective variation. In this paper, we proposed a multi-resolution enhanced cooperated localization method for the multi-type robots task. Based on multi-resolution iterative closest point (MRICP) and particle filter (PF), the proposed method could achieve a fast cooperation localization even without a given initial pose estimation. Firstly, a new registration evaluation function is introduced to value the registration efficiency and registration tendency. Secondly, to handle the posterior distribution problem in the PF based global localization, a map updater model is taken to control map scale, resolution level and the particles size to balance the registration efficiency and accuracy. Thirdly, a self-adoption particle resample scheme could deal with the sample impoverishment/degeneracy problem. Experiments with an unmanned aerial vehicle (UAV) and an unmanned surface vehicle (USV) are conducted for cooperate localization test at a water bay, and the results are analyzed to verify the feasibility and validity of the proposed method.