Machine Learning for Shared Control with Assistive Machines

The focus of this paper is to present a very relevant robotics application domain and its challenges, and to highlight how machine learning might be used to help resolve them. The domain in question is the shared control of partial automation assistive machines (e.g. powered wheelchairs, electric prostheses); where control is shared with the motor-impaired human user of the machine. A variety of challenges particular to this domain are outlined, for example appropriate control interfaces and user acceptance. A founding principle of the recently established Laboratory of Adaptive and Autonomous Rehabilitation Robotics is that machine learning can be used to address at least some of these challenges. This short paper elaborates on these ideas, identifying promising avenues for machine learning with the domain of shared control with assistive robots, and introduces a handful of projects within the lab beginning to address them. Keywords—Rehabilitation Robotics; Machine Learning; Shared Human-Machine Control