Food System Logistics Biomimicry - Ant Colony Optimization

Submitted by Casey W. Hoy on Mon, 02/20/2012 - 7:01pm

I was telling Kari Moore about this biomimicry-based technique for optimizing transportation routes, and figured the rest of this group might be interested as well, so I'm posting here rather than just emailing Kari.  Most of the work so far has been done in Europe, but we do have some OSU Fisher School of Business faculty who are beginning to work with this algorithm.  

Now for the musings of an entomologist - if you add the behavior of Argentine ants to the basic algorithm, you get a shifting nest site, and if you add in the optimal allocation of load size and distance to the body size of workers that a Malaysian leaf cutter ant species has been shown to follow, you'd have something even more like the challenge of local food system logistics - optimal vehicles and paths for multiple and ephemeral sources transported to multiple and varying markets or aggregation points.

Like a lot of things these days, there's a website devoted entirely to Ant Colony Optimization.  You can read more, probably more than you want to read at least in one sitting, at:  http://www.aco-metaheuristic.org/publications.html

The first overview pdf's are a good start.  I'd be interested in knowing if any of you are using this technique.

Best, Casey

 

Groups:

Comments

In my own experiments with

Submitted by Sam T. Rose on Sun, 03/04/2012 - 3:15am.

In my own experiments with Ants complex systems models, I've noticed that low evaporation rate, and high diffusion rate of modeled "phermone" chemical results in faster consumption of modeled "food". Lowe evap/high diffusion environment-modeled "ants" concentrate in swarms on "food" and clear out each "pile" of food faster because the swarm is able to concentrate and "find" the activity around each pile of food. 

 

Lowering the "evaporation" spreads too much spreads the signal out too much, so the ants run around looking for food that is supposed to be there, but is not. Raising evaporation too much doesn't leave enough signal for the next ant. Lowering diffusion too much isolates the signal too far away from the path ants are traveling. Swarming is *not* the classic application for ant colony simulation in solving shortest route transportation networks problem, but I still found it to be an interesting observation.