Use this URL to cite or link to this record in EThOS:
Title: Principles of sensorimotor control and learning in complex motor tasks
Author: Sylaidi, Anastasia
ISNI:       0000 0004 7228 4954
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2015
Availability of Full Text:
Access from EThOS:
Access from Institution:
The brain coordinates a continuous coupling between perception and action in the presence of uncertainty and incomplete knowledge about the world. This mapping is enabled by control policies and motor learning can be perceived as the update of such policies on the basis of improving performance given some task objectives. Despite substantial progress in computational sensorimotor control and empirical approaches to motor adaptation, to date it remains unclear how the brain learns motor control policies while updating its internal model of the world. In light of this challenge, we propose here a computational framework, which employs error-based learning and exploits the brain’s inherent link between forward models and feedback control to compute dynamically updated policies. The framework merges optimal feedback control (OFC) policy learning with a steady system identification of task dynamics so as to explain behavior in complex object manipulation tasks. Its formalization encompasses our empirical findings that action is learned and generalised both with regard to a body-based and an object-based frame of reference. Importantly, our approach predicts successfully how the brain makes continuous decisions for the generation of complex trajectories in an experimental paradigm of unfamiliar task conditions. A complementary method proposes an expansion of the motor learning perspective at the level of policy optimisation to the level of policy exploration. It employs computational analysis to reverse engineer and subsequently assess the control process in a whole body manipulation paradigm. Another contribution of this thesis is to associate motor psychophysics and computational motor control to their underlying neural foundation; a link which calls for further advancement in motor neuroscience and can inform our theoretical insight to sensorimotor processes in a context of physiological constraints. To this end, we design, build and test an fMRI-compatible haptic object manipulation system to relate closed-loop motor control studies to neurophysiology. The system is clinically adjusted and employed to host a naturalistic object manipulation paradigm on healthy human subjects and Friedreich’s ataxia patients. We present methodology that elicits neuroimaging correlates of sensorimotor control and learning and extracts longitudinal neurobehavioral markers of disease progression (i.e. neurodegeneration). Our findings enhance the understanding of sensorimotor control and learning mechanisms that underlie complex motor tasks. They furthermore provide a unified methodological platform to bridge the divide between behavior, computation and neural implementation with promising clinical and technological implications (e.g. diagnostics, robotics, BMI).
Supervisor: Faisal, Aldo Sponsor: Imperial College London
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral