26 Co-Evolving Painting Robots Using Graph Based Evolutionary Algorithms

Grid robots have proven to be an effective modeling tool for artificial life research, providing a platform to evaluate methods of machine learning. Some of the most interesting behaviors found in artificial life problems are the co-evolutionary effects caused by two different virtual species interacting with each other. This paper further examines this phenomenon by applying an artificial geography on two evolving species of virtual robots using evolutionary algorithms for controllers. Graph Based Evolutionary Algorithms were applied to three versions of a simple controller for “painter” robots. The results of this study showed that even though the virtual robot behavior changed, the preferred graph is consistent with the best performing graphs for a simple string problem. This verifies that the class of grid robot problems can occupy multiple regions of a taxonomy of evolutionary computation problems and allows for a baseline for future grid robot experimentation.