結合無線感測網路之機器人召喚系統設計

Abstract

本論文主要目的為設計一套基於無線感測網路之機器人召喚系統。當使用者召喚機器人時，機器人首先會透過Zigbee無線感測網路模組CC2431定位引擎來估測使用者之座標，並且將使用者之座標當作自主導航系統的目標位置，讓機器人朝向使用者前進。當機器人接近使用者時，透過Kinect感測器所提供的影像來偵測使用者人形與人臉，並且計算出使用者在影像中的位置資訊，接著透過影像追蹤控制器，讓機器人修正自身方向來朝向使用者前進，並且在設定的安全距離下，停在使用者面前。本論文所設計的機器人自我定位系統，整合里程計與CC2431定位引擎定位資訊，可以改善機器人因為移動過久、運作時間過長而產生的定位誤差，同時亦降低因無線感測網路室內定位之不確定性，使機器人可進行長距離之導航，找到使用者。經由實驗驗證，本論文所設計的機器人自我定位系統，在機器人移動60公尺後，機器人自身定位誤差平均為68公分。本論文所設計的機器人召喚系統，機器人能順利找到使用者，並且停在使用者面前與使用者互動。

In this thesis, a call-to-service system for mobile robots is developed based on a wireless sensor network. In this design, a ZigBee-based location-aware system is responsible for estimating the position of user who calls the robot for service. The autonomous navigation system takes the location information as goal position and guides the robot moving toward to the user upon calling. While the accuracy of the ZigBee-based location-aware system is limited, the robot uses Kinect to detect user's body and face to find the exact position of the user when the robot approaches the user. A visual tracking controller guides the robot to move toward to the user upon calling, and track the user in front of him/her by a set distance. For navigation control, we propose a self-localization system of a mobile robot by fusing the ZigBee-based location-aware system and odometer to improve the localization accuracy. Experimental results show that the average localization error is 68 cm in a 60 m travel. Experimental results also verified the effectiveness of finding the user upon calling at various distances and locations.