Oct. 19 (Thursday) Mike Sherback(Cornell University)
An Optimal Control Approach to Analyzing Sensorimotor Behavior
Abstract:
In this talk I will present the results of experiments testing the hypothesis that human sensorimotor response in a simple linear dynamical task can be accurately modelled in the classic LQG (Linear Quadratic Gaussian) framework. It is well known that the optimal controller for a linear system subject to additive gaussian noise is the LQG controller. By putting subjects in a situation with simple linear external dynamics and an additive gaussian disturbance, behavior largely consistent with the hypothesized optimality is observed. The process of fitting the results to LQG yields a parametrization of sensorimotor response in terms of motor noise intensity, loop delay, control cost weighting, and vision noise intensity that may have diagnostic utility.