Interactive Evolution of Ant Paintings S. Aupetit, V. Bordeau, N. Monmarch´ e, M. Slimane, G. Venturini Laboratoire d’Informatique de l’Universit´ e de Tours ´ Ecole Polytechnique de l’Universit´ e de Tours - D´ epartement Informatique 64 Av. Jean Portalis 37200 Tours - France monmarche,slimane,venturini @univ-tours.fr Abstract- In this paper, we present how we use an inter- active genetic algorithm to find the best parameters to build an artificial art work according to user’s aesthetic taste. Ants are used to spread colors on a numerical painting and behave with very simple rules to follow and deposit colors. These rules and colors are considered as parameters for the evolutionary process. This work can be considered as a contribution to naturally inspired ar- tificial art and evolutionary techniques are used to help artists in their creative process. 1 Introduction Naturally inspired algorithms are now well known and widely used such as genetic algorithms or artificial neural networks. More recently, ants have also inspired heuristics taking advantage of their collective and self organizing ca- pabilities (see [2] for an overview). In this paper we focus on the automatic production of numerical art works taking inspiration of ants’ behavior. As Nature have always been a predominant source of inspira- tion for art, its implication in numerical arts is shown here as the translation of an artificial life and artificial intelligence paradigm to produce a dynamic system able to create nu- meric paintings. We concentrate our study on two aspects: first, we will describe how ants can produce paintings with- out any human direction and second, we will introduce an evolutionary mechanism allowing users to interact with art work production. Automatic painting systems that use ants are still rare. For instance, in [19], the author describes ants that are able to pick up and deposit paint which represents food to pro- duce images and to analyze the emerging complexity and the regulation that appears in this kind of systems. With a greater artistic will, L. Moura and V. Ramos 1 have produced swarm paintings in a non human art approach. Other works consider an image as a playground for artificial ants and even for spiders but without any artistic goal, for instance to perform edge detection [15, 3]. Since our previous work on artificial ants deals with many different fields such as optimization [18, 13] or clus- 1 http://www.lxxl.pt/aswarm/fr_aswarm.html. tering [7, 11], artistic work produced by artificial ants is a new direction that we have already experimented with a sys- tem able to produce music with artificial ants [12]. The remainder of this paper is organized as follows: in section 2 we briefly describe the main principles of real ants’ behavior that we take into account. In section 3, we present how we use artificial ants to produce numeri- cal paintings. Interactive genetic algorithms are presented and applied to the exploration of parameters space for ant’s painting in section 4 and conclusions are drawn in section 5. 2 Real ants’ behavior Ants are social insects widely adapted to their various nat- ural environments [6]. Their communication abilities are often very complex and have inspired scientists, mainly to solve combinatorial optimization problems such as the clas- sical salesman problem [16]. The natural mechanisms that have been adapted are those which take place during mass recruitment of workers to find and exploit food resources. Different species are able to deposit olfactive substances, called pheromones, on the ground to guide workers toward food. This kind of communication is indirect because it re- lies on the ground and allows that auto-catalytic reinforce- ment mechanisms appear: the more ants are numerous to find a food site, the more they will deposit pheromones and the more other ants will be attracted by this olfactive path. Moreover, pheromones intensity decreases with time and in this way, ants can forget a path that leads to an empty site. This collective behavior has been modeled by a popula- tion of agents that take their decisions in a probabilistic way according to the pheromone intensities they encounter dur- ing their moves. These agents are simple in comparison to the complexity of the tasks they perform collectively. 3 Painting ants 3.1 Principles In this ant paradigm application, ant like agents can move on a virtual painting, which is a bitmap picture, laying down paint of different colors. At each step, an ant chooses the next pixel to reach in a pseudo-random manner: it has a