On Evolving Multi-Pheromone Ant Paintings

The use of swarm simulations and ant colony optimization methods for making abstract art has a brief but interesting history. We introduce some new techniques for evolving ant paintings using a non-interactive genetic algorithm. Key features of our virtual ant simulation model include: (1) the use of multiple pheromones, (2) the reliance on stigmergy to determine when an ant painting is finished, (3) ant interaction and exploration dynamics that lead to the formation of a background image and an overpainting, and (4) the design of a fitness function to guide the evolution of ant paintings subject to certain aesthetic goals. We show examples of our evolved ant paintings.