In studying the application of pulse-coded neural networks for bipedal locomotion, our biological model for neural organization is the network of neurons in the ventral horn of the spinal cord. This neural system meets our application requirements for involving large amounts of sensory input data, large numbers of coordinated output signals, and relative ease in determining whether or not our neural system is working. The ventral horn neural circuits were historically the first mammalian neural network systems to be studied in detail, and much of what we know about neural organization in the brain was first learned by studying spinal neuron organization. The context that underlies ventral horn neural organization can probably best be explained by first understanding what it is that these neural networks evolved to control: muscles. Even though any bipedal mechanical test platform we are likely to build at some point is ipso facto going to involve electromechanical actuators and sensors, an understanding of muscle organization is useful in defining the input/ output organization and servo dynamics of a robotic platform. In addition, by casting our neural network design in such a biomimetic framework we open the door to the possibility of future applications of what we learn about ventral horn networks to applications in neural prostheses. The purpose of this tech brief is to review the organization of muscles and muscle systems.
