本研究發展兩輪差速驅動機器人的視覺式同時定位與建圖(SLAM)系統。研究分為兩個部份:第一部份整合兩輪差速驅動機器人的軟硬體設備,架設視覺感測器與雷射測距儀,以VC++整合多種感測器與廠商提供的機器人驅動控制介面,以及使用雷射測距儀建立簡易的地面基準。第二部份使用三種方法估測兩輪差速驅動機器人的狀態,包括編碼器回授、影像回授、與混合影像與編碼器回授等方法。本論文所發展具備SLAM能力的行動機器人,在室內環境中成功測試路徑閉合的功能、執行長距離SLAM任務的能力、以及提供運動控制所需的機器人狀態回授。
In this thesis, the visual simultaneous localization and mapping (SLAM) is developed for differentially driven mobile robots. The research divided into two parts: First, the software and hardware system of the differentially driven mobile robot are integrated. We utilize a binocular vision and a laser ranger finder as the sensing devices for the robot. The sensing devices are integrated with the driving and control interfaces of the mobile robot using Visual C++ programming language. The laser ranger finder in this thesis is utilized to provide a simple method of ground truth. Second, the robot state estimator for SLAM is designed using three different kinds of feedback information provided by motor encoders, a free-moving camera and a camera with known motion model. The SLAM system is implemented on a differentially driven mobile robot. The basic capabilities of the SLAM system are successfully tested, including ground truth and loop closure, as well as the ability of navigating over a long distance in indoor environments.