Vetter, Philipp; (2001) Context estimation in sensorimotor control. Doctoral thesis (Ph.D), UCL (University College London).

Preview

Text Context_estimation_in_sensorim.pdf Download (10MB) | Preview

Abstract

Human motor behaviour is remarkably accurate and appropriate even though the properties of our own bodies as well as the objects we interact with vary over time. To adjust appropriately, the motor system has to estimate the context, that is the properties of objects in the world and the prevailing environmental conditions. This thesis examines how we estimate the context given that the context is nonstationary, it can change both deterministically and stochastically over time, and that information about the context, such as prior information and sensory feedback, may be incomplete or noisy. Psychophysical and brain imaging studies are described, showing that to determine the current context the central nervous system uses information from both prior knowledge of how the context might evolve over time and from the comparison of predicted and actual sensory feedback. A computational model is presented of how these two sources of information are modelled and combined within the central nervous system to derive an accurate estimate of the context that is then used to appropriately adjust the motor command selection. A modelling study explores how context estimation may be used to learn and select appropriate controllers within a modular architecture. The model suggests that prediction should precede control during motor learning and this is confirmed psychophysically. The experimental findings are discussed with regard to their computational significance and their possible neural representation.

Download activity - last month

Export as JSONXMLCSV

Download activity - last 12 months

Export as XMLJSONCSV

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item