Third day of the Summer School: Wednesday, June 8, 2010. The stars of the day have been Stephen Guerin and Owen Densmore. Their hands on demo on Agent Based Modeling convinced us all of the power of this technique. Not only to simulate behavior of ants, flocks or birds, but also to provide solutions for complex social and business problems.
In order to understand our world, and more important, to predict the future of our world, people have since long relied on modeling. Since the rise of low cost computing, modeling softwares have become mainstream. Latest type of modeling tools use the concept of agents, that interact with one another, in time and space. Agents can be anything, from ants, to birds, atoms, genes, humans...
The key in agent based modeling is that the logic is decentralized in each of the agents. This way the model comes a big step close to the real world.
Ants looking for food
The beyond example, using one of the dominant modeling softwares - Netlogo - models the behavior of ants looking for food.
In the demo, that you can see by clicking further in this text, we work with 250 ants. The environment consists of the ants nest, a food source and an obstacle between the nest and the food source. The ants look for the food, and once they find it, they leave pheromones on the ground to indicate the food trail to their mates. Curious how they work it out ? Click here!
Social and business applications
Ants looking for food: thats's nice. But what can agent based modeling do for society and business? Stephen and Owen had the answer. In the context of a large concert, they were requested to model the concert stadion and the concert attendees moving around in the stadion. This model proved to be the ideal tool to give the city's safety staff better insight in possible mass behaviors, in case of multiple unexpected events.
When I asked what had been done for business, the answer came quick and removed my latest doubts: project portfolio management! In a kind of bowling hall setting, we could see projects evolve, interact with one another, while the project portfolio manager was keeping an eye on the lanes, tipping from a beer in the lane seats! Wow, this I have to try, once home! I mean, the modeling of course, although ...
Comments
re: Santa Fe - Agent Based Modeling
By Paul Valckenaers 05/09/10 (2 months ago)
"The key in agent based modeling is that the logic is decentralized in each of the agents. This way the model comes a big step closer to the real world." It actually should be the reverse. A (software) system that stays closer to reality inherits some key properties of the real world (coherence, consistency). Most "enterprise systems" and most commercial simulation software have a poor design/architecture in this respect. They reduce activities to "documents, entities, ..." only to be surprised that these systems are inflexible, high-maintenance ...