Engage/Disengage: Control Triggers for Immersive Telepresence Robots

Abstract

Teleoperation encompasses the use of software and hardware interfaces to remotely control mechanical devices. Ideally, a telepresence humanoid robotic system, which faithfully replicates the operator's motion and accurately relays sensory feedback to the operator, should achieve tele-embodiment and reciprocally enable an immersive sense of presence for the remote operator. However, current research does not consider that operators coexist in a virtual and real environment and need continuous, alternative interaction with both virtual ('engage') or real world tasks ('disengage'). In this paper, we propose and explore the feasibility of an EMG-based control interface for disengagement from a fully immersive humanoid telepresence robot or virtual avatar - where the operator is manually and verbally constrained by the avatar's replication of the operator's motions. Our system makes use of a support vector machine (SVM) classifier, with 92.8% ± 2.5 accuracy on average, to continuously classify tongue gestures that indicate an engage or disengage command.