Neuromorphic Motor Control
The human motor control system involves many levels of control and is capable of learning new action combinations and sequences. In this project we model the operation of individual levels of this hierarchy, as well as interactions between levels. We use a neurally inspired message passing approach, using a generalization of factor graphs that allows for factorizations with exceptional dependencies, to quickly perform the inference needed to determine the currently appropriate motor action at each level. This approach will enable robust and responsive behavior in a real-time interactive environment.