2P1-F16 局所地図の時系列統合による大域地図の生成

概要

論文の詳細を見る
This paper describes a mapping method for mobile robots. Most previous works aim to make a precise map for a largescale environment. However, sensor information includes errors in general, and reliable information may not be obtained in some specific cases such as observing a wall with few textures by stereo. It is, therefore, difficult to make a global map directly from sensor information with large errors. We thus propose a mapping method which first makes a reliable local map and then stitches them into one global map. The method first makes a probabilistic, robot-centered occupancy map by Bayes inference using a sensor uncertainty model. It then makes a global map by applying a SLAM (simultaneous localization and mapping) method with considering the local maps as observation. The particle filter-based SLAM method used in this research needs estimates of robot motion in the prediction step, but the amount of calculation increases in case of less reliable estimates, for which the more particles are needed. We therefore estimate the motion by an ICP (iterative closest point) algorithm. The experimental results show the effectiveness of the proposed method.
一般社団法人日本機械学会の論文
2009-05-25