請用此 Handle URI 來引用此文件:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/76410
標題: | 具有情緒的自主社交陪伴機器人 Autonomous Social Companion Robot with Moods |
作者: | Zong-Ze Wu 吳宗澤 |
指導教授: | 傅立成(Li-Chen Fu) |
關鍵字: | 社交陪伴機器人;動態平衡理論;馬斯洛的五大需求;機器人自主系統;文本風格翻譯;視覺問題生成, Social Companion Robot;Homeostasis Theory;Maslow’s Hierarchy of Needs;Autonomous System for robot;Text Style Translation;Visual Question Generation, |
出版年 : | 2018 |
學位: | 碩士 |
摘要: | 隨著慢慢高齡化的社會跟勞力的不足,人們對社交陪伴型的機器人能陪伴老人或小孩的認知、意識逐漸增加,這是一個非常困難的問題,因為不論是老人或小孩都會更喜歡擁有一個有擁有自由意識的機器人而不是一個冷酷的機器人作為陪伴者。除此之外,為了使社交陪伴機器人能達到更高的自主性,機器人應該要能在沒有使用者的命令下自己做出決定。因此,我們利用穩態理論跟馬斯洛的五大需求來模擬我們的機器人,並架構了架構出自動系統來讓機器人知道每個時刻該做什麼。為了追求人與機器人能有更好的互動,在機器人與人在互動時某種程度上機器人該擁有自己的個性和情緒。與此同時,對於下一個階段的人機互動來說,人機互動不僅僅需要考慮機器人基本問答能力,也要將閒聊能力列入考量。因此,在本篇論文中,我們架構了一個能夠增進機器人與人之間的關係的聊天系統。另外,我們也開發了文本風格轉換的模型,此模型能根據機器人的情緒來轉換閒聊的神經對話系統的輸出句子。值得一提的是,好奇心是人的重要的特性之一,因此我們提出了一個視覺問題生成模型來讓機器人可以針對自己觀察到的景象提出多個問題。最後為了更好的去分析我們的系統,我們會去針對裡面各個提出的模組去做評估,例如我們的自動系統、文本風格轉換模型、視覺問題生成模型。
在實驗結果可以看出我們自動系統可以很好的讓機器人滿足自己的內在需求,並且在RSR指標達到99.98%,而在文本風格轉換模型中,我們的模型可以在Yelp的數據集中達到84.65%成功率來轉換文本的風格,除此之外,我們提出了幾種不同針對機器人的視覺問題生成模型的評估方式,而我們的模型經過評估後,是可以讓機器人有一定水準的互動。 With the gradually aging society and lack of labor, the awareness of social companion robots accompanying elders and children rises. It is a very challenging problem, since the elders and children might prefer having a robot with free wills rather than having a cold bloodless robot as friends. Moreover, for a social companion robot to reach high autonomy, it should make its own decisions without users’ command. Therefore, homeostatic theory and Maslow’s hierarchy of needs have been adopted to model the robot, and we build the autonomous system for robot to know what to do at every time moment. In order to pursue a better interaction between robot and human, it might be better for robot to possess personality and moods to some extent while interacting with human. Meanwhile, not only basic Question-Answering abilities but also abilities for Chit-Chatting should be considered crucial for the next level interactions. Therefore, in this research work, we build up a dialogue for improving the relationship between robot and human. What’s more, we develop a text style translation model to translate the style of the output sentence from chit-chat bot based on robot’s moods. It is quite worthy to mention curiosity, which is important characteristic of human, and thus we develop visual question generation capability such that robot knows how to propose diverse questions to human for what it has observed. Finally, we evaluate our system separately for the three proposed modules, such as the autonomous system, text style translation module, and the visual question generation module. The results show that the autonomous system can make the robot satisfy the internal needs well and reach 99.98% in RSR index. In text style translation, the transferred sentiment accuracy in the Yelp dataset can reach 84.65%, which is an improvement. With the proposed evaluation metrics, which is designed for the visual question generation of our robot, we can say that robot will have a certain quality for this interaction. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/76410 |
DOI: | 10.6342/NTU201802413 |
全文授權: | 同意授權(全球公開) |
電子全文公開日期: | 2023-08-09 |
顯示於系所單位: | 電機工程學系 |
文件中的檔案:
檔案 | 大小 | 格式 | |
---|---|---|---|
ntu-107-R05921012-1.pdf | 5.95 MB | Adobe PDF | 檢視/開啟 |
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。