Acquisition of Anticipatory Postural Adjustment through Cerebellar Learning in a Mobile Robot

Abstract

Anticipatory Postural Adjustments (APAs) are motor responses which anticipate a perturbation on the current body position caused by a voluntary act. Here we propose that APAs can be decomposed into a compensatory and an anticipatory component and that the cerebellum might be involved in the acquisition of such responses. To test this hypothesis, we use a cerebellar model to control the acquisition of an APA in a robotic task: we devise a setup where a mobile robot is trained to acquire an APA which minimizes a perturbation in its speed after a collision with an obstacle. Our results show that the same cerebellar model can support the acquisition of an APA separately learning its two sub-components. Moreover, our solution suggests that the acquisition of an APA involves two stages: acquisition of a compensatory motor response and prediction of an incoming sensory signal useful to trigger the same response in an anticipatory manner.