Continuous vocalization control of a full-scale assistive robot

Abstract

We present a physical robotic arm performing real-world tasks using continuous non-verbal vocalizations for control. Vocalization control provides fewer degrees of control freedom than are necessary to directly control complex robotic platforms. To bridge this gap, we evaluated three control methods: direct joint angle control of a selectable subset of joints, inverse kinematics control of the end effector, and control in a reduced-dimensionality synergy space. The synergy method is inspired by neural solutions to biological body redundancy problems. We conducted several evaluations of the three methods involving the real-world tasks of water bottle recycling and grocery bag moving. Users with no prior exposure to the system were able to perform these tasks effectively and were able to learn to be more efficient. This study demonstrates the feasibility of continuous non-verbal vocalizations for control of a full-scale assitive robot in a realistic context.