NRI: FND: The Robotic Rehab Gym: Specialized co-robot trainers working with multiple human trainees for optimal learning outcomes

University Of Wyoming, Laramie WY

Investigators

Abstract

This award will support research that will obtain new fundamental knowledge about how groups of robots can effectively teach different skills to groups of humans. While groups of robots can already assist groups of humans, existing research has mainly focused on short-term performance enhancement and/or safety and health. Instead, this project will advance the concept of a "robotic gym": multiple specialized robotic trainers that teach multiple different skills to groups of humans, with each robotic trainer focusing on different skills and each human learner having different initial abilities and different potential for improvement. This represents a long-term planning problem with significant uncertainty (especially since current performance cannot always predict future improvement), and will be addressed with advanced artificial intelligence techniques that will be evaluated both in simulations and in actual human subjects studies. Results of the research will advance national health and wellbeing in many ways. They can most directly be used for robot-aided training in motor rehabilitation, sports, and surgery; furthermore, they can be broadly applied to any setting where machines are used to train groups of humans, including education and weight loss. The team will also develop new interdisciplinary courses in human-robot interaction, and will perform outreach about robotics to multiple groups, especially K-12 and community college students and teachers around Wyoming. The project will be divided into three complementary work packages that will develop three elements of a proposed "robotic gym" - a multi-robot system that teaches multiple skills to groups of humans. The first work package will create an intelligent system that dynamically allocates multiple human trainees to multiple robotic trainers with the goal of maximizing individual and group training outcomes. The second work package will create ways of predicting future training outcome based on measurements of current and past performance, which can be used as a basis for decision-making by the intelligent allocation system. Finally, the third work package will create a collaborative human-robot planning framework that allows the robotic gym to learn training strategies from human experts (e.g., therapists) and to provide decision support to human experts, thus combining the complementary advantages of human and machine intelligence. The robotic gym will be evaluated in two ways throughout the project. First, the created software will be evaluated in simulations with many different variables (e.g., number of robots and humans, degree of uncertainty about current performance), providing the research community with detailed information about expected effectiveness in different scenarios. Second, it will be used to teach multiple challenging arm motions to groups of human participants using multiple robots available at the University of Wyoming, providing proof of feasibility for future applied research. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.