Applying digital evolution to the design of self-adaptive software

Abstract

As software developers, we strive to create computational systems that are as robust and versatile as biological organisms have evolved to be in nature. We propose a software development methodology capable of producing self-adaptive software, using digital evolution to discover behaviors and optimize solutions. Employing this methodology we present an example behavioral concept from inception to fruition on physical hardware, as a proof of concept of the approach. We evolve environmentally-aware motility behaviors through digital evolution, automatically translate the evolved programs into C code, and compile and load the programs onto mobile robots.