This paper aims to emulate human motion with a robot for the purpose of improving human-robot interaction (HRI). In order to engineer a robot that demonstrates functionally similar motion to humans, aspects of human motion such as variable stiffness must be captured. This paper successfully determined the variable stiffness humans use in the context of a 1 DOF disturbance rejection task by optimizing a time-varying stiffness parameter to experimental data in the context of a neuro-motor Simulink model. The significant improved agreement between the model and the experimental data in the disturbance rejection task after the addition of variable stiffness demonstrates how important variable stiffness is to creating a model of human motion. To enable a robot to emulate this motion, a predictive stiffness model was developed that attempts to reproduce the stiffness that a human would use in a given situation. The predictive stiffness model successfully decreases the error between the neuro-motor model and the experimental data when compared to the neuro-motor model with a constant stiffness value.

This content is only available via PDF.
You do not currently have access to this content.