Inter-Trial Dynamics In Goal-Oriented Tasks.

Open Access
Author:
John, Joby
Graduate Program:
Engineering Science and Mechanics
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
July 02, 2009
Committee Members:
  • Joseph Paul Cusumano, Dissertation Advisor
  • Joseph Paul Cusumano, Committee Chair
  • Karl Maxim Newell, Committee Member
  • Francesco Costanzo, Committee Member
  • Gary L Gray, Committee Member
  • Christopher Rahn, Committee Member
Keywords:
  • Variability in human movement
  • Performance
  • Inter-Trial Dynamics
  • Goal Equivalent Manifolds
  • Shuffleboard Experiment
Abstract:
Human beings can perform a variety of skilled tasks repeatedly and reliably. However, the details of joint trajectories and force production reveal considerable variability from one trial to the next. In this thesis we intend to understand the different factors that affect goal-level variability, a common measure of performance, in repeated trials of skilled tasks. Body-level variability has a significant effect on goal-level variability but it turns out that the body-level variability is just one of the factors; there are other passive and active factors that affect performance. We develop a novel theoretical framework to model the trial-to-trial dynamics arising in repeated trials. This approach, at its core, depends on the idea of a goal function: a mathematical definition of a task in terms of the subject's morphology, the physics of the environment, and the requirements of the task. This goal function, by definition must evaluate to zero for perfect performance. The set of possible body-states that satisfy this relation often has the structure of a goal equivalent manifold (GEM) on which the movement system attempts to converge while trying to achieve the goal. Our models help us analyze the ensuing dynamics and also demonstrates how the various factors like stability, noise and passive sensitivity interact with each other to produce performance. Using our approach we are able to develop data-analysis techniques that assess the stability of the inter-trial control process involved. Also, we are able to reconstruct the time-series of the controller action required to converge on the solution manifold and this sheds light on the structure of performance variability. To test some of the predictions of our model we develop a virtual-reality shuffleboard game. Based on our theory we predict the scaling of goal-level performance with respect to the passive sensitivity properties along the GEM. Using the understanding gained from our model, we also hypothesize that inter-trial dynamics is significantly more stable in a direction normal to the GEM while in the tangential direction it is weakly stable and that fluctuations in these directions are more uncorrelated (less persistent) and more persistent respectively. These hypotheses are tested and established to be true using the data-analysis methods we developed applied to the shuffleboard experiment. The theoretical framework developed in this work can be readily adapted to quantitatively characterize performance in a variety of skilled human tasks. We hope that the data analysis methods developed in this work can be used for movement analysis to distinguish healthy and pathological subjects and also to develop better training and rehabilitation techniques for athletes and people with various movement disorders.