Study of human forearm posture maintenance with a physiologically based robotic arm and spinal level neural controller.

The goals of this research are: (1) to apply knowledge of human neuro-musculo-skeletal motion control to a biomechanically designed, neural controlled, 'anthroform' robotic arm system, (2) to demonstrate that such a system is capable of responses that match those of the human arm reasonably well in comparable experiments, and (3) to utilize the anthroform arm system to study some controversial issues and to predict new phenomena of the human motion control system. A physiologically analogous artificial neural network controller and an anatomically accurate robotic testing elbow are applied in this study. In order to build the physical elbow system to have mechanical properties as close as possible to the human arm, McKibben pneumatic artificial muscles, force sensors, and mechanical muscle spindles are integrated in the system with anatomically accurate muscle attachment points. A physiologically analogous, artificial neural network controller is used to emulate the behavior of spinal segmental reflex circuitry including Ia and Ib afferent feedbacks. Systematic experiments of elbow posture maintenance are performed and compared with physiological experimental data. New experiments are performed in which responses to torque perturbation are measured when selected afferent pathways are blocked. A 'covariance diagram' is introduced. And a linear model is used to help to analyze the roles of system components. The results show that muscle co-contraction and Ia afference with gamma dynamic motoneuron excitation are two efficient ways to increase joint stiffness and damping, which in turn reduces the mechanical sensitivity of the joint to external perturbation and shortens the settling time of the system.