First try on the project

We did a first try with a flash animation to simulate an ant colony. The anthill is represented by a yellow square, the ants by a black one, the food is red and the obstacles are gray. The ants go randomly at first but they put pheromones on their way, and the other ants will most likely go for a "pheromoned" path. Once a food resource is found, the ant grab one unit of it and bring it back to the anthill following the pheromone path, and making it stronger.
Every time 10 unit of food are gathered a new ant is produced, which will look for more food. In this example two ant colonies are competing with each other, to allow us to compare the choice of different parameters. Here, the top left anthill is using less random parameters, and we can observe that it will generally loose.

We will then upgrade the project by having more ants and get rid of the squares to have a more natural look. Also we may try to use a genetic algorithm to find the parameters that lead to the most efficient and the most realistic behavior.

Presentation of aInts

We created this blog to present the progress of the aInts project. This project, which is part of the artificial intelligence course of Chalmers University, is to develop an ant colony algorithm. One of our inspirations for choosing this subject is the book "Empire of the ants" of the French writer Bernard Werber (more info: see http://www.bernardwerber.com/intro_en.html).

The book is a mix of science fiction, more or less accurate scientific facts and a passionating story. It describes an ant colony organization through a very interesting inside angle and gave us inspiration simulate such behaviors.

The concept of emergence also attracted us, about the magic of having lots of agents with very simple behaviors, and see complex and organized colony behaviors emerge out of it.

We are currently still looking for more information on the subject and thinking on how to narrow the project, but we currently want to focus more on simulation of the colony, having something that look like a real ant colony rather than trying to solve other algorithm problems with it.

The members of the project are:

Myriam Économou

Sonia Kostenko

Éric Hauchecorne

Raphaël Vandon