應用於室內自然環境且可與人自然溝通之居家服務機器人--子計畫三：居家服務機器人之即時動態導航與定位學習之設計與實現

Abstract

本計畫之主要目的在於研究居家服務機器人即時動態導航與定位之設計與實現，包 含：環境感知、導航(robot navigation)、以及迴避障礙等功能，使居家服務機器人能具有 自主認知與熟習環境的能力，以協助各個子計畫達成希望執行的任務。配合總計畫第 一、二年預定完成Robotcup@home 基礎及進階指定競賽項目的目標，本計畫將以：(1) 多感測器之障礙物偵測系統、(2)感測器資料融合、(3)虛擬環境地圖建立與定位、(4)自 主避障、(5)動態路徑規劃、以及(6)障礙空間資訊描述等六項研究重點，完成：(1)建立 一多重感知障礙物偵測系統，整合各種異質感測器之量測資訊，提供可靠之環境障礙物 資訊，(2)利用多種數據融合(sensor fusion)技術實現感測器資料融合，以提供可靠的量測 結果，作為建立地圖與避障之依據，(3)利用多種建地圖(map building)技術，建立並更新 未知環境之虛擬空間地圖，作為機器人自主導航的依據，(4)以多種啟引式演算法 (heuristic)以及演化計算法(evolutionary computation)實現最佳路徑之規劃，使機器人能據 以行動，快速且安全到達目的地，(5)依多重感測數據融合資料以及目標物移動方向之估 測，實現自主避障之功能，使機器人能安全行進，(6)利用多重感測數據融合資料作訊號 處理與座標轉換，實現障礙空間圖形化描述(obstacle profiling)，精確呈現障礙物空間資 訊。並與其他子計畫整合，使居家機器人能夠實現Robotcup@home 所指定之基礎及進 階競賽項目。第三年則針對適合台灣居家環境之機器人的相關項目進行研發，包含：(1) 提供居家空間多重輔助定位，包含超音波、RFID、以及影像式輔助定位，以校準機器人 位置，提供機器人精確定位，(2)建置即時遠端監控與整合式互動介面，方便使用者透過 遠端連線的方式進行監控，增加居家服務機器人的實用與便利性，(3)以嵌入式統軟硬體 協同設計(Hardware/Software Co-design)觀念，實現各種所提出之演算法，改善機器人定 位、路徑規劃、避障之執行效能，全面提升機器人之導航及避障性能。

The objective of this project is to investigate the design and implementation of dynamic navigation and self-localization systems for home service robots, including environment sensing, robot navigation, and obstacle avoidance, to equip the robot with the capabilities of autonomy for exploring the environment under consideration. Related sub-projects can benefit from the information provided by this project, including accurate position of the robot and optimal paths between the robot and the destination, in performing desired tasks set out in the Robotcup@home. According to the objectives to be achieved, the first and second years will focus on the preparation of primitive and advanced competitions of Robotcup@home. As a result, this project will investigate the following major items: (1) multi-sensing obstacle detection, (2) sensor fusion, (3) map buiding and localization, (4) obstacle avoidance, (5) dynamic path planning, and (6) obstacle profiling. It is hoped that, after the completion of the first two years, we can: (1) establish a multi-sensing obstacle detection system, integrating all of the heterogeneous sensors for providing reliable measurement, (2) use various sensor fusion techniques to provide reliable information for subsequent map building and obstacle avoidance, (3) use various techniques for map building to establish an environment map for an unknown environment, upon which robot navigation relies. (4) adopt A* and D* algorithms as well as evolution computation to derive optimal paths between the robot and the destination for safe and fast robot navigation, (5) based on the information obtained from sensor fusion and prediction of moving objects, implement a fuzzy obstacle avoidance controller, (6) obstacle profiling of the environment by processing results obtained from multi-sensor data fusion to accurately profile the obstacle environment. The objective in the third year of this project is to develop items related to the development of a localized home service robot suitable for use here in Taiwan, including (1) multiple auxiliary localization schemes, including ultrasonic-, RFID-, and image-based localization for providing extra position information for the robot, (2) implementation of the proposed algorithms for localization, path planning, and obstacle avoidance, etc, on a SOPC platform to improve the overall performance of the robot.