Manual control in a simulated pole balancing task: An intentional dynamics perspective

Date of Completion

January 2001


Psychology, Experimental




As a general framework for modeling goal-directed actions, intentional dynamics suggests an assessment strategy based on measuring the coupling between perceiving and acting during successful performance. This thesis investigates the relationship between information and control variables during performance on a simulated inverted pendulum control task. The information theoretic measure of mutual information is applied as a measurement strategy to identify those information variables that are relevant and actively involved in successful performance. In this exploratory research, mutual information is used to measure the relative influence of various information time series on a control time series under experimental conditions that manipulate physical parameters, boundary constraints, information availability, and task goals. ^ The results of the experiments suggest that the observables that are most relevant to maintaining control change over the course of task performance. In particular, the results suggests that in critical regions of the task space when failure avoidance behavior is required, the control strategy shifts from maintaining a stable dynamic to control action that is prospectively based on information specifying time and distance to a failure boundary. ^