Evolution of incremental complex behavior on cellular machines

Abstract

Complex multi-cellular organisms are the result of evolution over billions of years. Their ability to reproduce and survive through adaptation to selection pressure did not happen suddenly; it required gradual genome evolution that eventually led to an increased emergent complexity. In this paper we investigate the emergence of complexity in cellular machines, using two different evolutionary strategies. The first approach is a conventional genetic algorithm, where the target is the maximum complexity. This is compared to an incremental approach, where complexity is gradually evolved. We show that an incremental methodology could be better suited to help evolution to discover complex emergent behaviors. We also propose the usage of a genome parameter to detect the behavioral regime. The parameter may indicate if the evolving genomes are likely to be able to achieve more complex behaviors, giving information on the evolvability of the system. The experimental model used herein is based on 2-dimensional cellular automata. We show that the incremental approach is promising when evolution targets an increase of complexity.