Transforming plans into actions by tuning passive behavior: a field-approximation approach

The issue of transforming motor plans into actions as a field-approximation problem is considered. The approach is based on the idea that planning and control processes may use force fields for representing the interactions of a manipulator with its environment. On one hand, it is assumed that planning specifies targets, obstacles and other features as patterns of vectors over a manipulator's workspace. On the other hand, a manipulator operated by a set of impedance controllers acting in parallel is considered. Each controller modulates an output-force field according to a predefined law. The problem of setting the controller inputs is solved as a least-squares approximation of the planned vector field. The fields generated by the controllers are considered as basis fields, that is, as the vectorial equivalent of basis functions. In this context, the approximation problem is equivalent to finding which element of the functional space spanned by the basis fields is closest to the planned field. The relevance of this approach with respect to biological and artificial systems is discussed.<<ETX>>