Neural mechanisms for control of multivariable, redundant, nonlinear musculoskeletal systems

Engineering control theory exists to design state feedback that can cancel the nonlinear dynamics of a square multivariable plant and reduce the input-output relationship to a decoupled system of time-delays. When this theory is applied to the human musculoskeletal system two major difficulties are encountered. First, the system is highly redundant; some 700 functional muscles control some 110 elemental movements. Second, the central nervous system (CNS) has limited processing resources; by definition each elemental movement can be activated independently one at a time, but limited CNS resources prevent control of more than a few degrees of freedom simultaneously. This paper builds on previous work concerning the role neural adaptive filter circuitry (in particular that of the cerebellum) in self-organizing adaptive sensory-motor control. It discusses implementation of a biologically realistic controller that overcomes the above problems to produce energy- and resource-efficient synergetic movement