A study on nonverbal behaviors of humanoid robots

• Other Authors: Zheng, Minhua (author.)
• Format: Others
• Language: English; Chinese
• Published: 2015
• Subjects: Androids; Nonverbal communication; Robotics > Human factors; TJ211.49 .Z55 2015
• Online Access: http://repository.lib.cuhk.edu.hk/en/item/cuhk-1291490

As humanoid robots move from science fictions to reality, and are gradually being used in education, health care, and entertainment areas, interactions between humans and humanoid robots are becoming critically important. Previous findings show that in human-robot interactions (HRI) people tend to communicate with humanoid robots as if they were humans, which requires humanoid robots to be behaviorally more humanlike and socially more sophisticated. Non-verbal behaviors (e.g. gaze cues and gestures) are essential communication signals in human-human interactions (HHI), and they are equally important in HRI. This thesis reports a study on nonverbal behaviors of humanoid robots in hope to facilitate more natural HRI. Through extensive HHI and HRI experiments, intuitive robot gaze cues and gestures are studied, and their impacts on HRI are demonstrated. === Gaze cues can subtly mediate how human-human handovers take place. We conjecture that such effect also exists in human-robot handovers. Based on observations of the giver’s gaze behaviors during human-human handovers, several typical gaze patterns are extracted and transferred to a PR2 humanoid robot for carrying out robot-to-human handovers. In two consecutive HRI experiments the robot hands objects to human receivers while using different gaze patterns. Results show that where the robot gazes at and how it changes its gaze direction during the handover can significantly affect human receivers’ reaching time for the handed object and their subjective experience (likeability, anthropomorphism, etc.) of the handover. === Emblematic gestures are frequently used in HHI, because their meanings are self-contained and can be understood without spoken words, such as waving. We conjecture that emblematic gestures are also applicable to humanoid robots during HRI. Several commonly used emblematic gestures are identified and transferred to a NAO humanoid robot to be evaluated by human subjects. Results show that the perceived meanings of the robot’s emblematic gestures are generally consistent with the perceived meanings of a human’s emblematic gestures, but the recognition rate in the robot case is lower. To improve this situation, two design methods are implemented, i.e. by hand-puppeteering (designers manipulate the robot’s limbs with hands as if manipulating a puppet) and by motion mapping (human gesture trajectories are captured by an RGB-D sensor, and corresponding joint trajectories are mapped to the robot’s joints). Results show that gestures designed from the motion mapping method are faster and have larger range of motion, while gestures designed from the hand-puppeteering method are perceived subjectively as more likeable and as better conveying semantic meaning. === This research contributes to the design of humanoid robots’ nonverbal behaviors with theoretically and empirically grounded methodologies, and offers better understandings of gaze cues and gestures in both HHI and HRI. Findings from this research provide instructive and valuable references for many practical application scenarios involving interactive robots. === 人形機器人從科幻小說變成現實，逐漸被應用於教育、醫療、娛樂等領域，這使得人們與人形機器人之間的交互變得至關重要。之前的研究發現，在人-機器人交互中，人們傾向於以與人溝通的方式與人形機器人溝通，這就需要人形機器人具有更加似人的行為和社交類經驗。非語言行為（例如注視和手勢）在人-人交互中是必要的溝通信號，在人-機器人交互中這些行為也同樣重要。本論文致力於研究人形機器人的非語言行為，以期促進更自然的人-機器人交互。通過廣泛的人-人交互和人-機器人交互實驗，我們研究了直觀的機器人注視行為和手勢行為，並展示了這些行為對人-機器人交互的影響。 === 注視信號在人-人傳遞物品時能夠起到微妙的調節作用，我們推測這種作用也存在於人-機器人傳遞物品的過程中。通過觀察人-人傳遞物品時物品給予者的注視行為，我們提取了幾種典型的注視行為模式，並將這些注視行為轉化到一台PR2 人形機器人上，來進行機器人向人傳遞物品的實驗。在兩次人-機器人交互實驗中，受試者接收由機器人傳遞的物品，在此過程中機器人會使用不同的注視行為模式。實驗結果表明，機器人注視哪裡以及如何改變它的注視方向，能夠顯著影響接收者伸手接物品的時間以及他們對於傳遞物品這一事件的主觀感受（喜愛程度，似人程度等）。 === 象徵性手勢廣泛應用於人-人交互中，這類手勢具有獨立的含義、可以不依賴於語言而被人理解，比如揮手。我們推測在人-機器人交互中象徵性手勢也可為人形機器人所用。我們將幾個常用的象徵性手勢轉化到一台NAO 人形機器人上，並由受試者進行評估。實驗結果表明，人們對機器人的象徵性手勢的理解與對人的象徵性手勢的理解大致相同，但在機器人情形下的識別率較低。為了改善這種狀況，我們提出了兩種設計方法，即通過手把手操縱（設計者像操縱木偶一樣操縱機器人的肢體來獲得手勢動作），和通過動作映射（用RGB-D 傳感器捕捉人的手勢軌跡，將相應的關節軌跡映射到機器人的關節上）。實驗結果表明，通過動作映射方法設計的手勢速度更快、動作幅度更大，通過手把手操縱方法設計的手勢在主觀上更令人喜愛、更好地傳達了語義含義。 === 本研究為人形機器人的非語言行為設計提供了基於理論和實驗的方法論，有助於更好地理解人-人交互和人-機器人交互中的注視行為和手勢行為。本論文的研究成果對多種涉及交互式機器人的實際應用場合具有寶貴的指導價值。 === Zheng, Minhua. === Thesis Ph.D. Chinese University of Hong Kong 2015. === Includes bibliographical references (leaves 139-159). === Abstracts also in Chinese. === Title from PDF title page (viewed on 07, October, 2016). === Detailed summary in vernacular field only. === Detailed summary in vernacular field only. === Detailed summary in vernacular field only. === Detailed summary in vernacular field only.