Achieving Deployable Autonomy through Customizability and Human-in-the-Loop: A Case Study in Robot-assisted Feeding

Abstract

Despite decades of research on personal physically assistive robots for people with motor impairments, deployments of such robots are still few. Part of the reason is that every user's needs, environments, and care routines are unique, making it difficult to develop a sufficiently customized and robust robot. I present past and ongoing research with the ultimate aim of enabling a robot-assisted feeding system to feed a meal to any user, in any environment, without researcher intervention, in a way that aligns with the user's preferences. Our key insight is that the robot and user form a joint human-robot system that is working together to feed the user. Thus, we can achieve deployable autonomy by providing the user with intuitive and transparent controls to: customize the robot to their needs and environment; and make the robot's execution robust.