A Selective Vision and Landmark based Approach to Improve the Efficiency of Position Probability Grid Localization

概要

論文の詳細を見る
This paper presents a vision and landmark based approach to improve the efficiency of probability grid Markov localization for mobile robots. The proposed approach uses visual landmarks that can be detected by a rotating video camera on the robot. We assume that visual landmark positions in the map are known and that each landmark can be assigned to a certain landmark class. The method uses classes of observed landmarks and their relative arrangement to select regions in the robot posture space where the location probability density function is to be updated. Subsequent computations are performed only in these selected update regions thus the computational workload is significantly reduced. Probabilistic landmark-based localization method, details of the map and robot perception are discussed. A technique to compute the update regions and their parameters for selective computation is introduced. Simulation results are presented to show the effectiveness of the approach.
社団法人電気学会の論文
2003-04-01