2026-09-18T17:50:10Zhttps://publikationen.uni-tuebingen.de/oai/request
oai:publikationen.uni-tuebingen.de:10900/1502932025-05-22T01:06:02Zcom_10900_42126col_10900_42133
Publikationssystem UB Tuebingen advisor Kuchenbecker, Katherine J. (Prof. Dr.) author Mohan, Mayumi 2024-01-31T12:38:20Z 2026-10-09 http://hdl.handle.net/10900/150293 Die Dissertation ist gesperrt bis zum 09. Oktober 2026 ! The emergence of standalone exercise-coach robotic systems is an exciting prospect in the pursuit of reliable ways to help people learn new physical skills and maintain fitness. While socially assistive robots have demonstrated potential as exercise coaches, real-world deployment remains a challenge. Central to this issue is the topic of gesture-based interactions. When teaching or coaching, humans augment their words with carefully timed hand gestures, head and body movements, and facial expressions to provide feedback to their students. Robots, however, rarely utilize these complementary cues. A minimally supervised social robot equipped with these abilities could support people in exercising, physical therapy, and learning new activities. This thesis poses the question of how the intuitive power of human gestures can be harnessed to enhance human-robot interaction. In an effort to answer this query, the presented research explores gesture-based interactions to expand the capabilities of a socially assistive robotic exercise coach, investigating the perspectives of both novice users and exercise-therapy experts. This thesis begins by concentrating on the user's engagement with the robot, analyzing the feasibility of minimally supervised gesture-based interactions. This exploration seeks to establish a framework in which robots can interact with users in a more intuitive and responsive manner. The investigation then shifts its focus toward the professionals who are integral to the success of these innovative technologies: the exercise-therapy experts. Roboticists face the challenge of translating the knowledge of these experts into robotic interactions. We address this challenge by developing an algorithm that can enable exercise therapists to create customized gesture-based interactions for a robot, thereby bridging the knowledge gap. This thesis is thus divided into four parts, as follows. The initial part of this thesis focuses on the development of the Robot Interaction Studio, a cutting-edge, minimally supervised environment designed to facilitate social-physical interactions with an exercise robot. By leveraging markerless motion-capture technology, the system is capable of obtaining extensive quantitative information from a user as they interact with the robot. A within-subjects study with seven users was conducted to evaluate this setup, employing three distinct gesture-based interactions rooted in commonly observed social behaviors. Traditional exercise coaching systems often require heavy supervision by technical experts and can be cumbersome and time-consuming to set up, lacking the ability to provide sufficient information about user movement. This innovative system overcomes these limitations by generating rich, quantitative data that could enable more personalized interactions. The second part of this thesis shifts the focus to the robot's ability to provide meaningful feedback to the user. This phase also demonstrates how the Robot Interaction Studio can foster dynamic gesture-based interactions with a robot. Drawing inspiration from time-tested techniques found in the educational literature, we designed two distinct feedback paradigms for the robot, formative feedback and summative feedback, which are evaluated in the Robot Interaction Studio using the gesture-based cues developed in the first part of this thesis. Through a mixed-methods-design study with 28 participants, we demonstrate that robots can effectively utilize gesture-based feedback to enhance user performance and understanding, offering a novel approach to increase user engagement during physical activities with robots. The third section focuses on the professionals who are responsible for crafting exercise techniques for patients. A significant challenge is enabling the customizability of exercise-coach robots for use by exercise-therapy experts. Existing robots, often hard-programmed, cannot be easily modified by non-roboticists. We propose that motion-capture-based teleoperation could be used to resolve this issue by allowing an exercise expert to move their own body to create motions for a robot's arms and head. Thus, we utilize semi-structured interviews and qualitative surveys to understand the perspectives of eight experts on exercise robots and their teleoperation. We illustrate how infusing customizability through teleoperation can notably enhance experts' perception of robots, aligning the technology more closely with their needs. Lastly, the thesis turns to the development of an intuitive system that therapy experts could readily employ to create robot arm motions. Guided by the fact that experts are open to using teleoperation, the system is founded on motion-capture-based kinematic retargeting, an approach shown to be highly intuitive for human operators. An optimization-based algorithm named OCRA was created and thoroughly evaluated; the user can customize its behavior by setting one weight that dictates the relative importance of hand orientation error versus the shape of the arm in space, which we term arm skeleton error. It exhibits versatility across robots with varying limb lengths and degrees of freedom, provided they have revolute joints. A detailed study in which 70 users watched videos of a robot executing four operator motions each at four weight values demonstrates the algorithm's ability to replicate human-like motions. Rigorous benchmarking and validation further emphasize OCRA's robustness, making it a promising tool for creating robotic gestures. Thus, this thesis lays the groundwork for dynamic gesture-based interactions in minimally supervised environments, with implications for not only exercise-coach robots but also broader applications in human-robot interaction. en ubt-podok Gesture-Based Nonverbal Interaction for Exercise Robots PhDThesis URL https://publikationen.uni-tuebingen.de/xmlui/bitstream/10900/150293/2/DissertationMayumiMohanMAIN.pdf File MD5 a9c0db0c04c5fb47e07f18d14934fba5 36747305 application/pdf DissertationMayumiMohanMAIN.pdf URL https://publikationen.uni-tuebingen.de/xmlui/bitstream/10900/150293/3/DissertationMayumiMohanMAIN.pdf.txt File MD5 de9d522b6c3408563f3fa68039a2b6b0 411160 text/plain DissertationMayumiMohanMAIN.pdf.txt