On the neural network of a synthetic cerebellum to control fast movements of an arm

A synthetic motor control system which attempts to imitate or emulate the cerebellar function in movement control is studied. Some design considerations for such an emulator which would affect the neural network's architecture are discussed. Experimental results are given for dynamic compensation in the fast extension of a human arm based on a concept different from inverse dynamics. The observations illustrate how coordination of muscle activities can take advantage of mechanics of the arm and its dynamics to produce smooth fast movements