I. Introduction
Because of the richness and variety of its movements, control of the human arm by the central nervous system (CNS) has been studied extensively [1] – [3] Many studies have focused on the relation among the variables recorded from parts of the neuromusculoskeletal system such as the neural activity, muscle forces, and kinematic parameters. Others have proposed theories on how the CNS controls arm movements but none accounts well for the performance achieved despite the complexly nonlinear properties of both the sensors and the actuators. As a result, the task of restoring the lost functions to the paralyzed arm is very much like trying to repair a complicated system with little knowledge about its operational principles.