Evolution of Control System(s) for a Multi Joint Snake: Transformer <-> #13

Abstract

This thesis is about evolving a control system for a snake called Transformer <-> #13. This is a mechanical snake with several body parts. The choise was to use a cellular genetic algorithm where each body part is a cell. These contain DNA , one ruleset for each degree of freedom in the joints, which decides how it will behave in relation to its neighbour body parts. Three different fitness functions have been implemented which each gives a distinct and different behaviour. The goal of the different fitness functions is; crawling far, rising high and making geometry. The crawling part was successfull, while the other two goals was much harder for the snake and didnt provide great results. Concluding that the snake is appropriate for crawling around and making an impression of different cubic forms. Which for artist purposes is adequate, but it fails on getting into specific shapes.