Robotics has the potential of positively impacting quality of life, especially for people with special needs. If we are to meet the demand for personalized one-on-one care for the growing populations of elderly individuals and those with special cognitive and social needs throughout life, great strides must be made in human-robot interaction (HRI) in order to bring robotics into everyday application domains. This interdisciplinary project identifies a specific set of HRI research questions in socially assistive robotics, the study of robotic systems capable of providing help through social rather than physical interaction. The research foci of the study are: embodiment, personality, empathy, and adaptivity toward the development of an assistive HRI model for customized time-extended assistive interaction. The research will be grounded in the stroke rehabilitation domain, where personalized and dedicated care is needed to provide supervision, motivation, and training during the critical post-stroke period and beyond, and where assistive HRI can play a key role. Specifically, a novel assistive HRI model will be developed based on personality matching between the user and the robot, in order to optimize the user's task performance on rehabilitation exercises. The model will be evaluated on multiple testbeds with a large pool of human subjects from the stroke patient population. An online learning algorithm will enable the robot to adapt to the user both over the course of short-term interactions during a single therapy sessions (e.g., in response to mood and fatigue), and time-extended interactions over multiple therapy sessions (e.g., in response to the evolving recovery process over months of rehabilitation). The work is the first to study the role of personality and empathy in assistive HRI with human subjects, as well as to engage in longitudinal assistive HRI research to assess time-extended human-machine interaction in the assistive context. An important contribution of the research is the unified and tightly integrated end-to-end approach, which combines key HRI issues of embodiment, personality, empathy and adaptivity in hypothesis-testing experiments. Project outcomes will also include a large and unique corpus of multi-modal data, which will be collected and analyzed, and made available to researchers across the relevant disciplines. The scientific impact will go well beyond novel insights toward a better understanding of the fundamentals of assistive HRI, and the role and potential for assistive human-machine interaction for stroke patient populations and rehabilitation in general.
Broader Impacts: Currently there are about 750,000 new strokes per year in the United States, and some expect the number to double in the next twenty years with the growing elderly population. Project outcomes will provide pilot data necessary for translating the methodologies developed toward clinical applications.
The work on this grant focused on socially assistive robotics, intelligent non-contact robots that aid the user through social rather than physical interaction. The goal was to develop programs that would make socially assistive robots capable of aiding stroke patient rehabilitation. The work was motivated by the following challenges: 1) stroke rehabilitation can continue over a long time, even life-long, but only with dedicated task-oriented training (TOT); 2) TOT requires supervision and coaching, which is unavailable beyond the few insurance-covered occupational therapy sessions; 3) TOT for stroke patients is most effective in the course of daily life, involving the use of the stroke-disabled arm in typical daily activities (e.g., reaching for the coffee mug, shelving a book, etc.); and 4) using the stroke-disabled arm in daily activities can be demoralizing and depressing since it is slow and ineffective compared to using the other, healthy arm (but using the other arm means the stroke-disabled arm will not have the opportunity to recover). In summary, there is a great need for personalized coaching of stroke patients through task-oriented activities in daily life in order to promote rehabilitation. A research program was developed to explore how socially assistive robots can be used to serve as TOT coaches for stroke patients. We developed new technologies, including wearable sensors and new programs for socially assistive robots. We also evaluated the developed technologies in several studies with stroke patients interacting with the robots using different types of rehabilitation activities. Intellectual Merit: The work on this grant developed and tested a new wearable sensor that can track patient activity in everyday life. The information from the sensor can be used both by medical personnel for tracking recovery and, more importantly, by the patient, through user-friendly software or a robot, to monitor and encourage rehabilitation exercises. The research also developed an approach for taking the data from the sensor and using it to provide personalized challenges for the user based on their performance history and rate of improvement. The work on this grant also developed the first socially assistive robots for stroke rehabilitation, and demonstrated that stroke patients found interacting with those robots enjoyable and responded to the coaching provided. The robots used expressive movements, facial expressions, and speech to coach and motivate the patients. The results also found that the socially assistive robots were more effective and better accepted than computers providing the same type of rehabilitation exercises. The results of this grant include an approach to endowing robots with personality to match that of the user in order to be more effective coaches. Users who interacted with robots with personalities matching their own performed longer and better on the exercises, and also reported enjoying the robots more. Extroverted users were matched with extroverted robots that challenged them to exercise harder, while introverted users were matched by introverted robots that provided more praise and less challenge. Broader Impacts: There were over 800,000 new strokes in the US in 2012; that number will double in the next 15 years, due to the aging population and the obesity epidemic. The majority of stroke survivors have permanent disabilities that could be minimized with additional coached rehabilitation, but the scale of the problem makes human care infeasible. To the best of our knowledge, the work on this grant is the first to put stroke patients together with socially assistive non-contact personalized rehabilitation robots. Prior work has used non-social non-interactive robots to apply forces on stroke patient arms in order to promote recovery in the early stages after a stroke. This grant addresses the later stages, when the majority of patients no longer have professional coaching and rehabilitation; the goal was to develop a technology that can be provided after other existing in-clinic and outpatient care is exhausted, yet a great deal of rehabilitation is still needed. The resulting wearable sensors and robot programs can be used on a variety of different systems and products and present the opportunity for a new type of affordable and accessible rehabilitation technology that can be commercialized and made available to the vast and growing stroke-affected population The work on this grant has the potential to influence both near-term and long-term technologies for improved health and quality of life. To further broaden the impact of the grant, each year the research team members all got involved in large elementary and middle school presentations and used the robots to explain and demonstrate the purpose of their research and encourage the students to pursue studies in STEM (science, technology, engineering, and math).
