Jessica Marie Hatch with Roger Quinn, Dept. of Mechanical and Aerospace Engineering
Model of Neural Networks: Centralized Pattern Generator Comparator
Dr. Roger Quinn’s robotic’s laboratory is creating a robot controller capable of gait control and adaptation based on biological principles. In order to create mechanical analogues of biological systems, accurate physical and neural models are needed. These systems consist of a complex network of muscles, sensors and sensory neurons, interneurons, and centralized pattern generators (CPG). There is biological evidence to suggest that there is a spinal CPG (the origin of the desired signal) that works in conjunction with the output CPG (modulated by all the sensory signals and directly in control of the muscles) through a comparator element that makes corrections using a proportional-derivative type controller. In the simple case the model takes one signal input and compares that to the desired signal using the CPGs. The output of the comparator is the error between the two signals and corrections are then sent back to the muscles. The goal of this project will be to create a dynamic model the CPG Comparator by investigating combinations of control responses in order to optimize behavior and performance. Ultimately, we would like to have inputs at the low-level from multiple receptors and incorporate filters as part of the CPG Comparator.