Toward experimental validation of a model for human sensorimotor learning and control in teleoperation

Abstract

Humans, interacting with cyber–physical systems (CPS), formulate beliefs about the system’s dynamics. It is natural to expect that human operators, tasked with teleoperation, use these beliefs to control the remote robot. For tracking tasks in the resulting human–cyber–physical system (HCPS), theory suggests that human operators can achieve exponential tracking (in stable systems) without state estimation provided they possess an accurate model of the system’s dynamics. This internalized inverse model, however, renders a portion of the system state unobservable to the human operator—the zero dynamics. Prior work shows humans can track through observable linear dynamics, thus we focus on nonlinear dynamics rendered unobservable through tracking control. We propose experiments to assess the human operator’s ability to learn and invert such models, and distinguish this behavior from that achieved by pure feedback control.