Integrated hand/wrist control in a neuroprosthesis for individuals with tetraplegia

Independent grasp and wrist control in a hand grasp neuroprosthesis is compromised by moments at the wrist generated by grasp activation, gravity, and the presence of objects in the grasp. Since the interaction between hand grasp and wrist movement is somewhat predictable, a feedforward controller was designed with artificial neural networks to compensate for the grasp/wrist interaction. Computer simulations with a biomechanical model of the hand and arm were performed to evaluate the feedforward controller. The simulations revealed that in the absence of unpredictable disturbances, the feedforward controller was able to generate independent control of lateral hand grasp and wrist movement. An additional input to the wrist module representing arm orientation eliminated the errors in wrist angle due to gravity effects. Grasp and wrist errors due to unpredictable disturbances (e.g. large moments at the wrist generated by an object in grasp, muscle fatigue) could be reduced by the addition of voluntary wrist extension, and closed-loop feedback