Interaction Dynamics in Oscillator and Human-in-the-loop Systems.

Abstract

This dissertation addresses control system analysis and system identification in three areas: error propagation in synchronization of harmonic oscillators, modeling of human active movement, and identification of human control strategies in manual pursuit tracking. 1) While most studies of synchronization in oscillator systems have focused on the existence of synchronous solutions in steady state, many problems pertaining to the transient dynamics have not been fully resolved. We extend the well-established theory of fundamental limitations to study the transient error propagation (string stability) in a string of synchronized harmonic oscillators. We first translate design requirements in terms of time-domain response and hardware limitations into a set of constraints on closed-loop frequency response. We further capture the conflict between string stability on the one hand and time-domain design requirements and hardware limitations on the other through a new Bode integral. 2) Modeling human active movement is a challenging problem not only because muscle has very sophisticated and highly nonlinear dynamics but also because neural and other signals internal to the body are difficult to observe directly. We seek a simple yet general and competent model to describe active movement in object manipulation tasks. Inspired by the Norton equivalent circuit in electrical engineering, we build a model based on the motion and force/torque signals that may be observed at the points of contact between the human body and the environment. The model consists of a motion source to represent a human's motor plan and a spring-mass-damper coupler to capture the time-varying driving point impedance of the human hand. The model is validated using occasional experimental trials in which a participant experiences unexpected loads in a grasp and twist task. 3) Although a large amount of literature has provided methods to identify feedback control in manual tracking tasks, very little research has been undertaken to experimentally identify feedforward control. We capitalize on the theory of fundamental limitations to study the link between a human's ability to simultaneously reject disturbances and perform pursuit tracking. We further develop an identification method to separate human feedback and feedforward control strategies in sinusoidal tracking tasks.