Janan A. Smither - US grants

Thanks to advances in autonomy, an increasing variety of robotic devices have emerged over the last few years to assist disabled users with mobility and object manipulation. However, users often report higher satisfaction from controlling and interacting with a robot, even when an autonomous robot shows better quantitative performance, because they tend to see the robot not merely as an agent for retrieving objects but as a quintessential tool for reasserting their domain of interaction with their environment and engaging their available faculties to the fullest. Sadly, for these users effective interaction with a robotic device is frequently hindered by the fact that haptic feedback in the normal sense may not be possible, whether due to sensory and/or cognitive disabilities or to limitations imposed by motor disabilities that may prevent proper transfer of user intent in the absence of an appropriate user interface. Motivated by these considerations, and in the hope of boosting the current low rates of assistive technology adoption by its intended users, the PI will in this project create a software framework that is independent of the robotic system and allows for adaptively compensating the level and type of human-robot interaction to increase user satisfaction based on real-time measurements of user performance. Project outcomes will foster a new paradigm for the design of assistive technology, which will enable system developers to understand the impact of user preferences and incorporate them in their designs from the start, as opposed to the current inefficient practice of user testing a design after creating an expensive product or prototype.

This research will push the envelope of human-robot interaction through the creation of models for understanding the underlying intent of users with disabilities, which may adversely affect their environmental perception and response. By utilizing these empirical models to design an adaptive human-robot interface that can compensate for deficits and variability in user performance, the work will generate a novel framework for effective sharing of control between individuals with disabilities and their robotic assistants. Furthermore, the control design for physical human-robot interaction that will be developed as part of this work will advance the field of autonomous robotics in general through the creation of new algorithms for physically interacting with users and their environments. The research tasks include systematic modeling of user performance during human-robot interaction, estimation of user performance parameters within a generalized estimation framework, and design of adaptive Lyapunov-based control strategies to facilitate safe and efficient physical interaction of the robotic end-effector with the user and environmental objects. The work will be informed by quantitative/qualitative data to be gathered from extensive user studies in the field, by the PI's previous experience in working with this class of users, and by his expertise in assistive robotics.