外れ値処理を用いた確率的位置情報の統合による移動ロボットのロバストな位置推定

抄録

To realize robust localization of mobile robot, it is important to improve method of integrating position information from various sensors whose dependence on environment is different. In the case of probability localization, such as particle filter, simultaneous probability is used. However, when a sensor is affected by huge noises such as multipath of GPS, simultaneous probability is incorrect. The authors have proposed selective sensor fusion system based on similarity of distributions. This paper proposes a new similarity evaluation method analyzing particles sampled from distribution and verifies its effectiveness by real world localization.