Worker ants' rule-based genetic algorithms dealing with changing environments

Abstract

Contrary to popular belief, biologists discovered that worker ants are really not all hardworking. It has been found that in three separate 30-strong colonies of black Japanese ants (Myrmecina nipponica), about 20% of worker ants are diligent, 60% are ordinary, and 20% are lazy. That is called 20:60:20 rule. Though they are lazy, biologists suggested that lazy worker ants could be contributing something to the colony that is yet to be determined. This paper verified that genetic algorithms (GAs) with this worker ants' rule can solve an artificial ant problem efficiently in changing environments. In our approach, for each generation, we preserve not only individuals of high fitness but also individuals of low fitness. As a result of simulation conducted in a changing environment, the best performance of our proposed GA was obtained when the number of preserved individuals of high fitness and low fitness are each close to 20% of the population, while the remaining nearly 60% individuals are created by genetic operations, namely, crossover and mutation. This simulation result reinforces the 20:60:20 rule discovered in nature ant colonies. In a changing environment, this simulation result also indicates that worker ants' rule-based GA outperforms simple GA and CHC.