請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/60829
標題: | 應用於辦公室環境下基於社交感知之多人員意圖推論與機器人適應性決策機制 Social Awareness Multi-Human Intention Inference and Adaptive Robotic Decision Strategy in an Office Environment |
作者: | Yuan-Han Hsu 許元翰 |
指導教授: | 傅立成(Li-Chen Fu) |
關鍵字: | 社交感知,適應性決策,人機互動, Social aware,Adaptive decision strategy,Human robot interaction., |
出版年 : | 2013 |
學位: | 碩士 |
摘要: | 本論文提出一個應用於辦公室情境中,讓服務型機器人能自主性地推論出環
境中人員之社交意圖且動態地決策出適當的行為。由於機器人進入多人的辦公室 環境中,必須有能力推論出人員的社會行為或者意圖並自主性地在正確的時間與 地點提供合適的服務或反應機制。然而,推論出人類社交的規則是很艱困的。此 外人類的行為分析機推論出其心理狀態或意圖也是相當複雜的。幸運的是,心理 學家已在這近十幾年來中提出相當多針對人類社交行為的文獻與研究[1]-[5]。而 我們認為這些研究或許會輔佐我們的系統運作上更可靠且更有效率。 因此,在這篇論文中,我們採用社交信號處理與非語言的社交行為當作研究 架構中的特徵點,並利用雷射測距儀和RGB-D 感測器來偵測出人類的社交形態。 我們定義了人與人及人與機器人互動的社交模型,並同時提出社交感知之隱藏式 馬可夫模型(SAHMM)來推斷這些社交模式。另一方面,當機器人發現某些人員 需要服務時,則它必須做出適當的行為反應。因此,我們也必須提出了一個基於 SAHMM 架構下的社交意圖動態適應性決策模型。根據上述的理論,我們針對虛 擬與真實情境中進行了一連串的交叉實驗與理論驗證。而結果也符合我們預期 的,機器人能在社交環境中正確的推論出人類的社交意圖與執行適當的反應機制。 This thesis proposes a kind of service robots which can perform social-aware inference and dynamic decision service robots in an office environment. As the robots go into the multi-human office environment, it is important that they have ability to discover people’s social behaviour/intention and automatically provide proper services or reactions at the right time. However, discovering the human’s social interactions is tough. In addition, analysing people’s behaviours and then inferring their mentation or intention are also very challenging. Fortunately, the psychologists have been studying this work for several decades and proposing lots of theories about human social behaviours [1]-[5]. Those psychological theories could be able to support our framework to make it more reliable and more accurate. Thus, we would like to do the social signal processing with nonverbal social cues by employing laser range finder and RGB-D sensor to discover human social patterns in this thesis. Furthermore, we also model several social patterns including human-to-human, human-to-robot and multi-human-to-robot interactive formats. Besides, we also employ a Social-Aware Hidden Markov Model (SAHMM) to infer those social patterns in this research. After successful inference, the robots should try to implement the reactions appropriately when they discover some people who need services. Here, we propose an adaptive decision strategy based on SAHMM. On the whole, our system has implications for both human-robot interaction theory and design. The effectiveness of our proposed social-aware inference model and decision framework is verified with simulation and real testing scenario. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/60829 |
全文授權: | 有償授權 |
顯示於系所單位: | 資訊工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-102-1.pdf 目前未授權公開取用 | 5.07 MB | Adobe PDF |
系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。