### Tip of the Week #66                     Tip Index

Go to the Prior Tip Choice of Denominator in Investment Ranking Criterion
Go to the Next Tip Dow 36,000

# "Swarm Smarts"

## by Eric Bonabeau and Guy Théraulaz, Scientific American, March 2000, pp. 72-79.

### Solving Complex Problems Using Ant Models

As we moved into our new home in Los Angeles, in 1986, I carelessly put some food remnants into one of the shipping boxes in the yard.  The next day, I marveled at a continuous 2-way stream of ants forming a chain from their nest into the box.  The path of the ants appeared to optimized in a shortest-patch solution to a modestly complex geometry problem: straight from the nest to a point at the box base, then up at an angle to the nearest point of entry at the top. This was a living example of how nature often optimizes solutions.  It reminded me once researching in college how soap bubbles (shortest connections) and other physical systems can solve difficult problems.

"Swarm Smarts" provides an intriguing look at how modeling social insects is providing solutions to several daunting problems.  Ants are simple creatures, whose behavior can be described with simple rules. I once read an article about programming a robot to mimic the foraging behavior of an ant on a beach.  This article shows how a community of simple-acting agents can work together to accomplish complex tasks.

When an ant leaves the nest looking for food, it leaves a pheromone (scent) trail for it to follow back and for other ants to follow.  The first foraging ant making it back to the next leaves a double-trail (one trail leaving, and one returning). Thus, the shortest route is doubly marked, and more ants will follow it.  This simple model finds the shortest route between the nest and a food source.  Allowing the pheromone trail to "evaporate" (as in nature) provides the ants a mechanism to explore for alternate food sources when the first is depleting and for alternate routes should the first become blocked.

By analogy, businesses want to solve problems such as:

• The "traveling salesman problem," where the shortest route is desired to visit a series of cities.  Similar problems include the shortest or least-congested connection path for Internet traffic and lowest-cost sequence for drilling wells in an oil field development.  The ant-like models provide robust solutions with automatic backup plans.
• Looking for patterns in data.  "Data mining" is the hot topic in database management.  Similar to ant behavior in organizing a nest (e.g., sorting dead into piles), data can be sorted by placing like objects near other like objects.   The number and types of clusters emerge automatically from the data.
• Letting a swarm of slightly intelligent object work together to solve complex problems.  Ants will work together to transport a large food object back to the nest.   "Dump parts, properly connected into a swarm, yield smart results."

This is a fun article about how a biological can solve human business problems by analogy. For 3-1/2 decades Scientific American has been my favorite scientific journal—including economics, the "dismal science."

—John Schuyler, February 2000