RESEARCH

The real-time control of a complex musculo-skeletal system such as the human body requires the generation of thousands of control signals per second. Humans' ability to accomplish difficult tasks - in the face of noise, delays, uncertainty, and constantly changing circumstances - suggests that these control signals are chosen rather intelligently and to a large extent online. Our goal is building a computational theory of the sensorimotor loops responsible for this moment-to-moment control and testing the theory experimentally. The right framework for such a theory appears to be stochastic optimal control on the motor side and Bayesian inference on the sensory side; the two are dual in a deep mathematical sense. The specific projects in the lab fall in several categories: developing efficient control algorithms suitable for biomechanical systems; constructing control-theoretic models of behavioral phenomena; testing model predictions in motor psychophysics experiments; running exploratory experiments in the absence of predictions; implementing systems-level models in recurrent neural networks.

SELECTED PUBLICATIONS BY TOPIC   (full list here)

REVIEWS

Mathematical introduction to optimal control theory (2006)
Optimality principles in sensorimotor control (2004)

CONTROL THEORY

Linearly-solvable Markov decision problems (2006)
Hierarchical feedback control of redundant manipulators (2005)
Iterative LQG methods for constrained nonlinear stochastic systems (2005)
Optimal control and estimation with state- and control-dependent noise (2005)

MOTOR BEHAVIOR

Evidence for the flexible strategies predicted by optimal feedback control (2006)
Analysis of the synergies underlying complex hand manipulation (2004)
Unsupervised learning of sensorimotor primitives (2003)
Optimal feedback control as a theory of motor coordination (2002)

MOTOR PHYSIOLOGY

Optimality of mixed muscle-movement representations (2006)
On the role of primary motor cortex in arm movement control (2003)
Cosine tuning minimizes error and effort (2002)
Direct cortical control of muscle activation (2000)

RELEVANT CONFERENCES

Advances in Computational Motor Control (ACMC)
American Control Conference (ACC)
Computational and Systems Neuroscience (CoSyNe)
Engineering in Medicine and Biology (EMBC)
Neural Control of Movement (NCM)
Neural Information Processing Systems (NIPS)
Society for Neuroscience Meeting (SFN)