Learning Control Sharing Strategies for Assistive Cyber-Physical Systems

Abstract:

People with upper extremity disabilities are gaining increased independence through the use of assistive robotic arms, but performing tasks that require many small precise movements remains difficult. In fact, a confounding factor is that the more severe a person’s motor impairment, the more limited are the control interfaces available to them to operate the system. One technique for reconciling the mismatch in input and system dimensional complexity is to implement multiple control modes, that is multiple subsets of the control dimensions. However, switching between control modes is expensive; our own interviews with current users of a Kinova arm pinpointed that the struggles with modal control relate back to the need to constantly change modes. In this work, we present a formalism for assistive mode switching that is grounded in hybrid dynamical systems theory. The robot can anticipate user mode switches and perform the switch automatically. Additionally, we describe a formalism for human and robot mutual adaptation, where the robot guides the human teammate towards more efficient strategies, while maintaining human trust to the robot.