臺灣大學: 電機工程學研究所傅立成李俊毅Li, Jiun-YiJiun-YiLi2013-03-272018-07-062013-03-272018-07-062011http://ntur.lib.ntu.edu.tw//handle/246246/253957這篇論文提出了一個基於情境感知的智慧型機器人導航，隨著機器人應用的增加，未來將有越來越多的機器人進入我們的生活空間，在沒有預先得知周遭環境的情況下，機器人必須可以感知周遭環境的情境進而做出相對應的行為。首先，我們會定義一些設計的標準來規範機器人的行為，接著我們會利用雷射測距儀來偵測環境和人所造成的情境。在環境所造成的情境方面，我們會依照雷射測距儀的資訊來判斷周遭環境的特性，例如鄰近的障礙物或牆壁，在人所造成的情境方面，我們會利用人員追蹤系統來偵測機器人周遭的人，依照人的資訊我們可以判斷人對機器人是否有和機器人互動的意圖，機器人在導航的過程中不會一直將人視為障礙物，在人有互動意圖的時候可以停下來與人互動。為了確保人的個人空間不會被侵犯，我們應用了人類敏感場的概念，此敏感場的大小是根據人與人的關係及其心理因素所決定的，並且可以將人的感受考慮進來。此外，在人群的情境中，我們提出了一個基於晶格圖的短程路經規劃技術，對於人位置的預測，我們使用擴展卡曼濾波器(EKF)來獲得更準確的結果，利用些改過的 搜尋法可以找出最佳的路徑，使機器人可以先前進到比較安全的地方進而找出機會穿越人群。最後，本論文也會透過在多人環境中的模擬及實驗來驗證情境感知導航的可行性。This thesis proposes a new navigation framework for mobile robot to maneuver in an unknown and populated environment. Without being given the knowledge of the surrounding environment, the robot has to use reactive mechanism to deal with unforeseen circumstances. To achieve this, we develop a context-aware navigation scheme that enables the robot to perceive the current context of the environment and make an appropriate reaction. The robot will actively detect humans and extract geometric features by only using a single laser range finder. So far, many rules have been proposed or formulated based on human''s appearance to extract human information. In this thesis, after obtaining the human information from the human tracking module, the robot is capable of determining human''s interaction intention. Humans are no longer treated as obstacles all the time during the navigation process. Instead, the concept of Human Sensitive Field is adopted in order to ensure intactness of the personal space, the robot will take human feeling into consideration when it encounters the human. Furthermore, the robot will make an appropriate action in accordance with the context it encounters. In particular, to deal with the situation when the robot encounters crowds, we develop a short-term path planning technique using lattice-based graph search while applying Extended Kalman Filter to obtain more reliable human motion prediction in this situation. Based on the proposed Context-Aware navigation, the robot is able to navigate through a multi-human environment in a smooth and socially acceptable manner.5610414 bytesapplication/pdfen-US情境感知以人為本人機互動Context-AwareHuman-OrientedHuman Robot Interactionthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/253957/1/ntu-100-R98921007-1.pdf