智慧型機器人於動態社交環境下以人為感知之互動行為

Human-aware Interaction Behaviors of an Intelligent Robot in a Dynamic Social Environment

曾士桓(Shih-Huan Tseng)

指導教授 ： 傅立成

國立臺灣大學/電機資訊學院/資訊工程學研究所/博士(2016年)

https://doi.org/10.6342/NTU201603478

摘要

在機器人開始與一群人互動的情況下,需要解決這些的問題“這 群人在哪裡?”和“機器人什麼時候應該接近他們?“,因此本論文開 發了一個新的系統,使機器人識別當前的社交情境是什麼情況後確 定何接近上述之人群,與它們進行互動。該系統主要是融合深度相關 的數據來追蹤這群人的位置,通過深度相關的數據中提取這些人的 社交提示,以及使用動態決策網絡(DDN)模型適時提供服務。主要 的挑戰在於機器人需先了解人群的社交提示進而了解機器人和人群 之間的社交情境。社交提示是包括人群的人際距離學 (Proxemics) 和 交談時的隊形 (F-formation),而社交情況基於社交提示被分成個別交 談 (individual-to-individual)、個人對機器人 (individual-to-robot)、機器 人對個人 (robot-to-individual)、人群對機器人 (group-to-robot)、機器人 對人群 (robot-to-group)、講悄悄話 (confidential discussion)、團體討論 (group discussion)。我們的系統執行如後所述: 機器人一旦發現有一群 人後,開始提取目標人群的社交提示,適當推斷相應的社交情境,然 而機器人決定是否應該發起基於指定規則的互動行為。所進行的實驗 結果表明該系統設計的適當性和所提出的方法的有效性,對於辨識人 群表現的社交提示以及考慮到機器人與人群的社交情境。
要完成從一個機器人提供更加自然和智慧服務,機器人不僅能夠按 照使用者的意圖和執行情況理想的服務,同時也感知用戶的反饋,需 要滿足用戶的期望下,調整服務內容和/或執行。故在這篇論文中,提 出了一項搭配使用者自然語言回饋以及含有推理策略的學習計畫,使 得決策網絡和動態貝氏網絡具有適應性,以滿足人類使用者的期望。 實驗結果表明,所提出的學習計劃,使機器人的任務執行有效地自主 適應使用者的期望。此外,我們證明了此系統足以讓人滿意的性能, 也就是系統成功地推理出人的意圖,以及機器人決定出使用者期望的任務。

關鍵字

人機互動 ； 初始互動 ； 以人為感知的動態馬爾可夫決策模型 ； 使用者意圖 ； 任務規劃 ； 強化學習。

並列摘要

In a situation where a robot initiates interaction with a group of people, questions such as ”where is the people group?” and ”when the robot should approach them?” should be addressed. This thesis develops a new system that enables a robot to determine when it approaches the aforementioned human group and interact with them after identifying what the current social situation is. The system is mainly to fuse depth-related data to track the positions of a group of people, extract social cues of those people by using depth-related data, and a dynamic decision network (DDN) model to provide service in proper time. The main challenge lies in understanding of the social cues of the group and the current underlying social situation concerning the relation between the robot and the group. The social cues are consist of Proxemics and F-formations, whereas the social situations based on social cues are categorized as individual-to-individual, individual-to-robot, robot-to-individual, group-to-robot, robot-to-group, confidential discussion and group discussion. Our system proceeds as follows : once a group of people are detected and the social cues of that target group of people are extracted, the corresponding social situation is appropriately inferred, and in turn the robot decides whether it should initiate interaction with the group based on rules to be specified later. The conducted experimental results demonstrate the properness of the system design and the efficacy of the proposed method in recognizing the social cues among individuals of the group as well as the nature of the social situations concerning the group and the robot.
To accomplish more natural and intelligent service provided by a robot, the robot should not only be able to perform the desirable service in accordance with the user intention and the circumstances, but also perceive user feedbacks and adjust service content and/or executions if needed to meet the user's expectation. In this thesis, the learning scheme with a reasoning strategy is proposed to adapt the dynamic decision network. The robot can analyze a user's natural speech feedback to adjust its decisions and the current social situation as what the user expects through the strategy. The experimental results show the effectiveness of the proposed learning scheme that enables autonomous adaptation of robot's tasks to fill the user's expectation.

並列關鍵字

Human-robot interaction ； initiate interaction ； human-aware dynamic decision networks ； data fusion ； social situations ； reinforcement learning.

參考文獻

[2] Edward T. Hall. The hidden dimension, 1966.

[3] Aaron Steinfeld, Terrence Fong, David Kaber, Michael Lewis, Jean Scholtz, Alan Schultz, and Michael Goodrich. ”Common metrics for human-robot interaction”. In 1st ACM SIGCHI/SIGART Conference on Human-Robot Interaction, volume 30, pages 33–40, 2006.

[7] N.Mitsunaga, Z.Miyashita, K.Shinozawa, T.Miyashita, H.Ishiguro, and N.Hagita. ”What makes people accept a robot in a social environment - discussion from six- week study in an office”. IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 3336–3343, 2008.

[9] Ambady Nalini, Frank J. Bernieri, and Jennifer A. Richeson. ”Toward a histology of social behavior: Judgmental accuracy from thin slices of the behavioral stream”. Advances in experimental social psychology, 32:201 207, 2000.

[10] M. Pantic nd H. Bourlard Vinciarelli. Social signal processing: Survey of an emerg- ing domain. Image and Vision Computing, 2009.