基於深度影像資訊之救災探勘機器人協同系統的設計與研發(I)

Abstract

近年來將機器人應用在救災方面的設計是一個相當熱門的應用，其中如救災搜索機器人的研究， 主要是可以在因地震倒塌的建築物內決定搜索地點。此類機器人可以記錄位置資訊，辨識已搜索和尚 未搜索的地方。將機器人應用在救災方面的設計的好處是可以有效且快速的進行受困民眾的搜尋以及 可以透過機器人的探索了解倒塌的建築物的內部情形以方便救難人員進入搜救，有鑑於此，本計畫書 提出了一個基於深度影像資訊之救災探勘機器人計畫，提供在倒塌的建築物中搜尋受困者的功能並且 協助搜救人員進入救援受難者的功能，本計畫主要包含3D場景重建、機器人定位、機器人可行路徑判 斷以及路徑規劃透過3D場景重建可以知道倒塌的建築物內部的情況藉以幫助救難人員了解倒塌的建 築物內部情況，而機器人定位則是幫助理解機器人行進的方向、角度以協助建立路徑，機器人可行路 徑判斷有助於機器人朝正確的方向前進，最後再透過路徑規劃規劃出適合搜救人員進入的最短路徑。 如果多隻機器人可進行探勘，則可藉著彼此之間的資訊傳輸與協同合作，避免相同區域重複探勘，進 而增加探勘搜索效率。

Recently, the design of rescuer robots has become a popular research field. Among these researches, rescuers-searching robot is mainly used to decide the search spot in the collapsed building. This kind of robot can memorize the information of location and identify the explored or unexplored region. One advantage of applying robots to rescue is that robots can search the victims effectively and efficiently. Another advantage is that the inner condition of the collapsed building can be understood through the exploration of robots, so as to aid rescuers enter to take action. As a result, this proposal proposes a study of 3D depth-image based rescue and exploration multi-robot system. It provides function of searching the trapped people and helping rescuers enter to relive the victims. This project mainly includes 3D reconstruction、localization of robot、motion planning of robot and path planning. By 3D reconstruction, rescuers can be informed of the inner condition of collapsed building; however, localization helps us record moving direction and angle of robots and build the paths. Motion planning helps robots moving toward the right direction and finally mapping out the shortest path suitable for rescuers to enter through path planning. If many robots are used, re-exploration for same section can be avoided by information transmission and cooperation among robots, and the performance can be improved effectively.