全向式移動機器人之同步定位與環境地圖建立

Abstract

本論文提出一使用全向式攝影機(Omni-directional camera)之全向移動式機器人(Omni-directional Mobile Robot)之同步定位與環境地圖建立(Simultaneous Localization and Mapping, SLAM)。藉由全向式攝影機360°的視野，取得更豐富的影像資訊和更多的環境地標(Landmark)，並結合擴展式卡門濾波器(Extended Kalman Filter, EKF)達成全向式移動平台的vSLAM。本論文採用參考點轉換與建立的策略，讓機器人能自主建立個別參考點，當走回已到過區域時仍能從資料庫中取回儲存之參考點資訊使用，減少EKF的計算負擔。實驗結果顯示特徵點辨識之正確率為93%，以實驗室之機器人於室內繞行8字形軌跡移動行經約16公尺後之起點與終點定位誤差均在0.1公尺以內。實驗結果證實機器人能藉定位系統達成走廊上的長距離全向式移動及轉動，且建立走廊環境的特徵點地圖，達成全向移動式機器人的室內導航功能。

This thesis investigates simultaneous localization and mapping (SLAM) of an omni-directional mobile robot. A method is proposed to use an omni-directional camera to realize SLAM algorithm based on extended Kalman filter (EKF). This study focus on the use of 360° of view of the omni-directional camera to reduce the accumulative error from odometer and to achieve simultaneous localization and mapping of the omni-directional platform. The method of visual reference scans is adopted in this design. Features of previously visited places can be used repeatedly to reduce the complexity of extended Kalman filter. Experimental results show that the matching rate of landmark features is 93%. The localization error is less than 0.10m for traveling 16 meters of “8” shaped route. Indoor navigation experiments revealed that the proposed localization system can navigate the omni-directional robot in an indoor environment and build the feature map simultaneously.