The exact role of the cerebellum in motor control and learning is not yet fully understood. The structure, connectivity and plasticity within cerebellar cortex has been extensively studied, but the patterns of connectivity and interaction with other brain structures, and the computational significance of these patterns, is less well known and a matter of debate.
In this project, we compared two existing models of the role of the cerebellum in error-driven motor adaptation – one in which the cerebellum generates feedforward corrective motor commands (feedforward model) and one in which the cerebellum acts in tandem with the brainstem through a recurrent loop (recurrent model). We found striking differences between these models in the expected patterns of adaptation to different kinds of disturbance.
Specifically, we have identified two distinct classes of sensorimotor disturbance – one corresponding to a change in the mapping between motor command and observed outcome and one corresponding to a change in the mapping between the stimulus and the desired outcome. Examples of both of these kinds of mapping can be found in vestibulo-ocular reflex (VOR) adaptation, saccadic gain adaptation and in adaptation of reaching movements. The exent and rate of adaptation varies significantly across architectures and across classes of disturbance in all of these settings.
This work offers some insight into the computational implications of various different proposed cerebellar architectures. As well as being of importance for cerebellar modelling, it may be of use in biomimetic and bio-inspired robotic control systems, as we have demonstrated with a hybrid architecture for VOR control which, although not particularly plausible biologically, is able to combine the benefits of and eliminate the weaknesses of the two proposed cerebellar architectures, providing robust control and adaptation under all conditions.
Related Publications and Presentations
- Adrian M Haith, Carl Jackson, Chris Miall, and Sethu Vijayakumar, “Unifying the sensory and motor components of sensorimotor adaptation”, Advances in Neural Information Processing Systems (NIPS), Vancouver, 2009.
- Adrian M Haith, and Sethu Vijayakumar, “Robustness of VOR and OKR adaptation under kinematics and dynamics transformations”, IEEE International Conference on Development and Learning (ICDL), 6th, 2007.
- Adrian M Haith, Carl Jackson, Chris Miall, and Sethu Vijayakumar, “Interactions between sensory and motor components of adaptation predicted by a Bayesian model”, Advances in Computational Motor Control, 2008.
- L. Paninski, Adrian M Haith, and Gabor Szirtes, “Integral equation methods for computing likelihoods and their derivatives in the integrate-and-fire model”, Journal of Computational Neuroscience, 2007, 10.1007/s10827-007-0042-x.
- L. Paninski, Adrian M Haith, Jonathan W. Pillow, and Chris Williams, “Improved numberical methods for computing likelihoods in the stochastic integrate-and-fire model”, COSYNE, 2005.
- Adrian M Haith, and Sethu Vijayakumar, “Implications of different classes of sensorimotor disturbance for cerebellar-based motor learning models”, Biological Cybernetics, 2008.