This work describes a method of generating gaits for the open-loop control of robots using genetic algorithms.
While similar techniques have been used before on different robots, the focus of this work is in controlling hyper redundant snake-like robots for which motion planning is much more complex than wheeled or tracked robots.
A dynamic rigid body simulation models a simple version of both the robot and its environment, and is used to evaluate the fitness of a proposed gait.
A genetic algorithm uses the simulation in order to rapidly test different gaits, and evolves new gaits using the fitness data from the simulation.
This system allows the rapid and automatic development of new and better gaits, without time consuming tests on a physical robot, or human intervention.
It is hoped that the method described will prove useful in creating gaits for complex robotic morphologies, which defy human attempts of control.