An Optimal Control Approach to Sensorimotor Parametrization

We tested the hypothesis that human sensorimotor response can be modelled as a linear quadratic gaussian (LQG) controller (a Bayesian optimal state estimator in series with a linear quadratic regulator), thereby inferring parameters characterizing human sensorimotor response in a simple experimental context. Subjects used a modified computer mouse to attempt to keep a displayed cursor at a fixed desired location despite a gaussian random disturbance and simple cursor dynamics. Data were fit to an LQG model whose assumptions are simple and generally consistent with other sensorimotor work, greatly reducing the dimensionality of time series data in a way that preserved the dynamic qualities of the system while yielding a small number of parameters. The parameters lend themselves to physical interpretation: noise intensity, control cost weighting, and delay. Inferred control cost and motor noise intensity showed statistically significant variation across subjects. The quality of the fit and of the inferred parameters generally support the hypothesis.