Email updates

Keep up to date with the latest news and content from BMC Neuroscience and BioMed Central.

This article is part of the supplement: Sixteenth Annual Computational Neuroscience Meeting: CNS*2007

Open Access Poster presentation

Anticipative adaptive muscle control: forward modeling with self-induced disturbances and recruitment

Christoph Kolodziejski1*, Bernd Porr2 and Florentin Woergoetter13

Author Affiliations

1 Bernstein Centre for Computational Neuroscience, University of Goettingen, Bunsenstr. 10, 37073 Goettingen, Germany

2 Department of Electronics & Electrical Engineering, University of Glasgow, Glasgow, GT12 8LT, UK

3 Department of Psychology, University of Stirling, Stirling FK9 4LA, UK

For all author emails, please log on.

BMC Neuroscience 2007, 8(Suppl 2):P202  doi:10.1186/1471-2202-8-S2-P202

The electronic version of this article is the complete one and can be found online at:


Published:6 July 2007

© 2007 Kolodziejski et al; licensee BioMed Central Ltd.

Poster presentation

A motor system must be able to adapt to perturbations in a fast and robust way. Additionally, in the long run learning leads to motor skills which, for example, allow humans to move the forearm with varying speeds. A generally excepted view on motor learning includes forward models that are generated on top of an existing control loop provided by reflexes. Through closed-loop feedback control a much improved motion sequence "without thinking" can be executed and later be adapted to changes in the environment. In this study we show that it is possible to combine temporal sequence learning with compliant joints and antagonistic muscle control to learn a forward model of a reflex.

We also show that the model is executed with the required strength, which depends on self-induced perturbations or external forces. Here self-induced perturbation means, for example, the change of torque by accelerating the forearm. For this purpose we apply a learning rule paired with recruitment which we have recently introduced. The rule correlates the mono-synaptic reflex loop provided by each muscle with an anticipative control signal which is used to move the upper arm segment of a two-joint arm. Without learning the deviation is compensated after some delay. After learning the antagonistic muscle pair stiffens and immediately reacts to the self-induced disturbance with the required muscle force. With this simple recruitment mechanism it is also possible to compensate different forces, caused by varying upper arm accelerations, without delay and re-learning. This kind of learning creates a forward model of the already existing mono-synaptic feedback loop and we are able to show that applying learning with recruitment to a compliant motor control structure quickly compensates self-induced disturbance.

Acknowledgements

We would like to give acknowledgments to the support from PACO+.