Evolution in Natural and Artificial Systems

The goal of this research is to study the conditions in which evolution may lead to the sustained emergence of novel behaviours, and how this may be applied to the automatic design of complex entities. We argue that with regard to artificial evolution, this field of study has been rather overlooked, with a preference given to mathematical or experimental results based on abstract selection models. We argue further that when practicioners of artificial evolution have tried to harness the dynamics of more natural selective processes, this relative lack of connection with traditional evolutionary theory has often led to implicit assumptions and “commonsense” intuitions, which may have prevented a better understanding (and more reasonable expectations) about these systems. Based on a study of evolutionary theory, we expect to provide reasonably clear definitions of what we may expect to observe in evolutionary experiments. We will put these ideas to test through the realisation of a software platform which will allow us to simulate the evolution of threedimensional creatures interacting in a physically realistic world. This software platform will place both morphology and control of the creatures under evolutionary control, thus allowing them to construct their own behaviours in response to their environment. By choosing different settings for competition, selection, reproduction and representation of the creatures, we will try to discern which conditions allow interesting behaviours or evolutionary dynamics to emerge.