Co-evolution of Configuration and Control for Homogenous Modular Robots

Modular robots are well suited to implement key features of autonomous machines such as versatility, adaptability and reliability. Our vision is to tackle this task following the three major axes (phylogeny, ontogeny and epigenesis) that underlie the emergence of autonomous and self-organizing organisms in nature. This paper presents Adam, our modular robot simulation and evolution tool. Adam successfully implements the first step of our project, inspired by the phylogenetic axis. We co-evolve configuration and control of locomoting homogenous modular robots by means of genetic programming. A tree-based genotype is used, encoding the control as well as the configuration of the modules. The modular robots are evaluated in a simulator that accurately models rigid body dynamics. Furthermore, we propose a grammar for an intuitive script that allows building modular robots ‘by hand’ or inspecting and manipulating evolved individuals.