System Identification of Human Dynamics Using Periodic Impedance Models During Physical Interaction

Human impedance models have emerged as a powerful framework for describing behavior based on the kinematic (generalized displacement/velocity) and kinetic (generalized force/torque) relationship, and such models have become essential for the rehabilitation sciences, and the development of advanced prosthetic and orthotic devices. However, given that most techniques include a perturbation-based protocol for system identification, a major challenge is extending the impedance modeling framework to more natural and realistic settings. In this paper, we introduce an impedance model with periodic terms for the stiffness and virtual trajectories. This allows us to fit the model without perturbations during periodic tasks. Using a single-degree-of-freedom forearm rotational haptic manipulandum, we fit our model to N=10 participants during dynamic physical interaction in four different virtual environments. Our model provides an excellent fit (average R 2 = 0.99) to the underlying data with minimum parameter variation across participants suggesting the periodic impedance model can accurately capture human dynamic behavior.