Comparing Adaptivity of Ants using NEAT and rtNEAT
Mario Sanchez and Teo Gelles
Creative Commons CC BY 4.0
Although individual ants have an extremely basic intelligence, and are completely incapable of surviving on their own, colonies of ants can develop remarkably sophisticated and biologically successful behavior. This paper discusses a set of experiments which attempt to simulate one of these behaviors, namely the ability of ants to place pheromones as a way of communication. These experiments involved a variety of different environments, and tested two varieties of the genetic algorithm NEAT: the standard offline version, and its online counterpart rtNEAT. Since the experimental environment did not seem to offer any benefit to continuous learning, we had expected NEAT and rtNEAT to have roughly similar learning curves. However, our results directly contradict this hypothesis, showing much more succesful learning with rtNEAT than with standard NEAT.