Projects

Robotic Teleoperation to Autonomy through Learning

Abstract

The goal of this research is to create a human-robot collaborative system that learns from sensor-assisted teleoperation to reach the maximum possible autonomy requiring minimal user input. It is assumed that the user while unable to do a task manually, can use a haptic interface to perform it in teleoperation. This system will utilize (a) the cognitive and perceptual abilities of an individual in a wheelchair with significantly limited motor skills, AND (b) the superior dexterity, range of motion, and grasping power along with intention recognition and learning capability of a robot. Our system will seek autonomy by continuously learning from sensor-assisted and imprecise teleoperation by a human. Initially, the workload will be distributed between the robot and human based on their abilities to perform ADL/IADL related basic actions in (i) autonomous or (ii) sensor assisted teleoperation modes. Any tasks or portions of the tasks that can be performed autonomously will be done in autonomous mode and the remaining tasks or subtasks will be done in sensor-assisted teleoperation mode. The human-robot collaborative system will gradually learn to perform these tasks autonomously in the long run.

Motivation

Individuals with diminished physical capabilities must often rely on assistants to perform Activities of Daily Living (ADL) or other Instrumental Activities of Daily Living (IADL). Even though lacking or severely limited in gross and fine motor skills, such individuals are often in possession of sound cognitive and perceptual abilities. In this work, we will focus on individuals with Muscular Dystrophy (MD), Multiple Sclerosis (MS), and Spinal Cord Injuries (SCI levels C5 to C7).