The Evolution of Complex Attributes in a Species of Simulated Agents

Abstract

Translate Article

In order for evolution to populate the planet with multiple species, two processes need to be at work. One is speciation, which involves the development of a new reproductively isolated species from an ancestral one. The other is that a reproductively isolated species can evolve to be more complex and potentially more capable over time. This second process is what we address in the research reported in this paper. One of the issues with developing individuals with more complex attributes is that those attributes typically take multiple mutations to be viable and before that point, the changes made by mutations are often detrimental. In this research, we use a simulated environment filled with simple agents and a genetic algorithm operating on these agents, each of which has its own set of chromosomes. We use this to test the plausibility of the species developing more complex structures (in this case sight) as the genetic algorithm population survives and evolves over 1000s of generations. In order to better simulate intermediate detrimental mutations, each mutation results in an increased metabolic cost for the agent, leading to higher energy expenditure with every movement or action taken. Tests were done with differing number of mutations required before sight was developed. The results showed that evolution of complex attributes are possible depending on the number of detrimental mutations required to make the attribute an advantage to the agent.