Note: this description contains vocabulary specific to this project, if you find yourself lost, refer to the Vocabulary Page, which covers the definition of terms for the group seminar.

The simulation is actually a number of simulations, each aimed at exploring a different aspect of the virtual world. Each has some unique quality, but all have the same basic framework.

The "basic" simulation (the one pictured on the main page) is the base of the project. Different explorations will be described later.

The simulation begins by building an environment. The environment is made up of an 18x18 grid of patches. Each patch can have some properties, in this case the basic property is food. Food is represented by green, the more intense the green, the more food present. At the begining, each patch has a 1 in 3 chance of starting with food. The amount of food produced at that time is also random. So, at the begining, about one third of the patches have food, and the amount of food on each of those patches is random.

The environment is then populated with a number of Decision Making Turtles (DMTs). They are named turtles, not for their appearance, but because of their roots. Much of the terminology and many of the concepts used in this project owe a great deal of credit to many whose work precedes my own. These works include Turtle Graphics, Logo, and more specifically StarLogo. More information on StarLogo (an artificial simulation environment), and the people behind it, can be found at http://el.www.media.mit.edu/groups/el/Projects/starlogo/home.htm.

All DMTs begin with the same amount of energy, the same speed and are placed in a random location in the environment. Each time they move they lose some energy. Each time they eat they gain some energy. The amount of energy the DMT loses when it moves is based on it's speed, so a fast moving DMT will lose more energy than a slow DMT. But, a fast DMT may be able to find large food sources easier than a slow DMT. The act of eating from a patch will deplete that patch's food. Black patch's hold no food.

Each patch also has some random chance of re-generating food each frame. Even patches that did not start with food can still re-generate food, so all patches can support life. The chance that a patch will re-generate food is based on the current number of live DMTs, so that large populations will have more food available to them than small populations.

Each DMT begins with a random genome, which is what it uses to make decisions (See Decision Making Procedures for more information). Some DMTs will be better at surviving, some worse. Few will survive through the first generation. A DMTs decision making genome can not change during it's life time.

This population of diverse individuals, with their decisions based on a simple genome, is a prime candidate for evolution by genetic algorithm (See Evolution/GA Procedures for more information). The genetic algorithm will select the better performing DMTs from the population, create offspring and place 72 new DMTs in the environment, with the same initial conditions of energy, speed and location. Patches again generate a random configuration of food and the simulation is restarted.

Each new generation will, hopefully, improve the survival rate of the DMTs. Because this procedure is based in part on random selections, some generations may do slightly worse than the previous, but overall the population will improve over time.

View this simulation, and a couple others similar to it, in the Demos section.