Computational human sensorimotor control.

The main objective of this programme is to study the computations underlying planning and learning in human sensorimotor control. The programme will investigate three projects. Dynamic motor learning: The advent of robotic technology that can generate computer controlled force-fields that perturb movements has led to a dramatic increase of our understanding of sensorimotor learning. In this project we will develop novel robotic technology which will allow realistic simulations of virtual hand-held objects to examine dynamic learning in tool use. Probabilistic mechanism in sensorimotor learning: Sensory and motor uncertainty form fundamental constraints on human sensorimotor control. This project will further develop our work on Bayesian sensorimotor estimation and optimal control, focusing on how the CNS deals with uncertainty. Statistics of action: The statistics of sensory inputs, suchas natural scenes or sounds determine the way the CNS develops and represents the world. However, there is no data on the natural statistics of everyday movements and this project will be the first to examine the relationship between the natural statistics of actions and sensorimotor control processes. The overall goal is to integrate these three inter-related areas so as to provide a cohesive understanding of the computational processes involved in sensorimotor control.