多機器人系統之遠端合作感測與環境模型建立

Abstract

環境地圖是智慧型機器人導航重要的資訊，如何建立環境地圖一直是個受到重視的研究領域。本論文提出一個運用多機器人做合作式感測的架構，以單一攝影機及物體與地面的交線特性擷取環境特徵影像，建立以線段為基礎的二維環境模型。以機器人配備之USB攝影機，我們發展出以影像處理及攝影機校正計算物體之距離資訊，並利用高斯分布建立量測的不確定程度，透過卡曼濾波器的概念融合來自不同機器人的資料，建立較為精確且完整的環境地圖。本論文以Java語言實現以網際網路建構之遠端呈現與機器人操控介面。經由整合實驗，顯示所提出之方法確有預期之功效。

This thesis presents a method to construct a two-dimensional(2D) map based on range data gathered from multiple mobile robots. The scenic range data is obtained from single camera equipped on the robot by using baselines of objects and image processing. The 2D map is represented by multiple line segments. Line segment parameters obtained from different mobile robots are fused to get a more precise 2D map and with reduced uncertainties utilizing Kalman filter techniques. The constructed map can be accessed remotely at a local site and the robots can be controlled through Internet. The functions on the server and client side are realized using Java language. Integrated experimental results show that the system has expected performance.