The central research goal of this work was to design and evaluate hardware and algorithms that can personalize to infant therapy needs and motivate motion practice. During the course of this work, we designed and evaluated an assistive robot and new robot modules; designed and evaluated a behavior-tree-based software architecture for informing the robot; completed short-term experiments with human-robot playgroups; and conducted long-term studies with the assistive robot. This work was conducted in conjunction with the Eugene, OR Child Development and Rehabilitation Center (CDRC).

Experiments conducted during the early parts of this work showed that when our custom assistive robot (GoBot) was added to a playgroup or one-on-one interaction with children with typical development, the children tended to move more. There seemed to be something different and extra motivating about the robot, even though other developmentally appropriate toys were also present in all experiment conditions. This differential advantage of socially assistive robots compared to other types of toys and technologies is consistent with past related results in child-robot interaction research.

We also performed iterative human-centered design of our robot to better customize it to the needs and interests of children with disabilities. These design cycles were followed by pilot experimentation and then complete long-term experiments considering the interactions between our robot and children with motor disabilities. In these evaluations, as with the studies with children with typical development, the participants tended to move more when the robot was active in their play space.

Additional research products emergent from the project included modular hardware for GoBot. Specifically, we created light, basic sound, bubble, ball launching, air dancer, button, and music modules for the robot. Over the course of iterative design, we identified and implemented safety improvements for the robot. On the software side, we formulated and evaluated a unique behavior tree architecture for robot decision-making. New sensing techniques were needed to enable the project. Our team designed and evaluated an overhead camera-based method for tracking child motion. We also designed and evaluated custom software for "tag" and "keep-away" interactions with the robot.

Over the course of the grant, we trained seven graduate students (four PhD and three MS; one from kinesiology and the remainder from robotics graduate programs) who contributed to the project work. Fifteen undergraduate students (7 REUs, 8 part-time researchers; four from kinesiology/health programs, and the rest from engineering). The interaction between team members in engineering and kinesiology supported the successful cross-disciplinary exchange of information and perspectives. Likewise, the team’s interactions with the CDRC resulted in mutually beneficial learning and discourse. The team collectively produced six peer-reviewed academic journal papers and eight peer-reviewed academic conference papers based on the project work. An additional journal paper is currently under review.

The products of the work overall can contribute to child wellness. Others with interest in similar topics can build off of the hardware and software resulting from the project. The insights from the work, which showed recurring potential for an assistive robot to increase child motion levels, can inform future physical therapy and early motion encouragement strategies, in addition to potentially sparking promising new clinical trials on the same topic.

Last Modified: 05/25/2025
Modified by: Naomi T Fitter