CAREER: Towards Programmable Social Robots for Everyone: A Teacher-in-the-Loop Learning from Demonstration Framework

Oakland University, Rochester MI

Investigators

Abstract

Social robots have demonstrated significant positive impacts across a range of application areas including delivering K-12 education, therapy for autism spectrum disorder, and older adult care. Although they are intended to provide much needed support in these areas, social robots remain inaccessible to individuals (e.g., healthcare professionals, educators, and service industry workers) who could best use the technology to produce the greatest impact to our society. This is because social robots still require significant technical expertise to be deployed and there is currently no clear pathway for laypeople to utilize social robots for their desired applications. This research will bridge the gap between social robots and individuals with no robotics/programming experience so that social robots can be made accessible to everyone. This will be accomplished by developing approaches and technologies that empower laypeople to teach and customize social robots for their desired applications, thereby transforming our emerging healthcare and education workforce by increasing their capacity for delivering services previously not possible. Project outcomes will also broadly impact education by supporting interdisciplinary workforce development and broadening participation in STEM for individuals from a range of backgrounds. To address these goals, this project will investigate fundamental approaches and techniques that enable a layperson to teach a social robot via demonstration. Teaching is often an iterative process that requires a teacher and learner to coordinate their efforts towards the mutual goal of knowledge transfer but there currently is minimal effort in engaging a layperson during the robot teaching process beyond providing data or examples the robot can learn from. This research will keep teachers-in-the-loop throughout the entire robot learning process by: 1) developing interfaces that enable a teacher to prepare high-quality demonstrations, 2) developing approaches that improve a teacher's understanding of models learned by a robot, and 3) developing approaches that enable teachers to integrate domain knowledge while they teach a robot. Project outcomes will both advance knowledge on enabling laypeople to independently teach a social robot and our fundamental understanding on how humans transfer social knowledge to machines. 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.