How Can Ant Pheromone Behavior be Simulated in a Computer Program?

[mXSCNT]

I want to simulate some "ants" on a computer, as a cellular automaton. I'm not going for realism, but I want to capture the ability of ants to find paths between nests and food sources, using pheremone information (stigmergy), and purely _local_ observations. I've been working on a program for this but I'm having difficulty getting the pheremones to work.

My attempts have resulted in a few interesting images, such as this one:
http://img24.imageshack.us/my.php?image=bluerock.tif

But my attempts are not successful at leading the ants to food. I'm wondering what I'm doing wrong, and hoping some biology will aid me.

Does an ant use multiple kinds of pheremone for leaving trails?

If you have a pheremone trail running north-south between food at the north and a nest at the south, and you put an ant on the trail facing south, will it turn around to go for the food, or will it follow in the same direction it was set down? (in other words is there something in the pheremones to indicate direction or is it just intensity?)

If it is just intensity, do ants derive direction from the intensity, by following the pheremone gradient? Or do they just go more or less straight while staying close to the pheremones?

From how far away can an ant detect food or smell pheremone trails? How far do ants travel from the nest in search of food?

Do foraging ants try to get close to (or far away from) other foraging ants, or do they only care about the pheremones?

Do ants constantly drop pheremones to get others to follow, or only after they have found food?

 

[jim mcnamara]

No direction per se, but ants accumulate more pheromone molecules on the shortest path to food. So every path has to 'remember' how many of your ants traversed it.
And ants have to 'smell out' the most heavily used path.

Google for 'ant colony algorithm' or 'adaptive pheromone path planning'