Media Coverage
September 2010
Concluding the Ant Trilogy
We’ve discussed Ant Colony Optimisation more than any reasonable person should. So, to close off the subject once and for all, I thought I would point you in the direction of “Smart Swarms” – a book by senior National Geographic journalist Peter Miller that covers the subject in far more detail than we can ever hope to cover in this column – as well as mention a very specific case to highlight the impact it can have.
Based on his research for an earlier article, the book describes Miller’s interactions with a number of biologists working with swarm-like animals in nature: bees, ants, pigeons and fish schools, amongst others.
What they told him was that large natural groups have little to no individual intelligence guiding them, but used a series of instinctive rule-sets that governed their behaviours in ways that were incredibly beneficial to the group as a whole. Many of their problems are decentralised ones, yet they are adept at solving it in a decentralised way.
Humans likewise work best when in groups, but our individual intelligence precludes being able to viewing very similar problems – such as those found in the complex environment of logistics - in a decentralised way.
Part of the problem is we lack the numbers and the liberal nature of a swarm. Most logistic companies are dwarfed by your average ant colony, and despite what your accountant may say about the Operations department, their specialised knowledge means giving them a driver’s cap and sending them out on deliveries won’t improve matters.
We have a tendency to rein in (with limited success) complexity through centralisation, rather than allow that complexity to shape our behaviours. We favour top-down organisational structures, and often decisions which have far-reaching implications throughout a company are made by a small number of individual egos at the boardroom level.
New leaps in software and online communication are, however, helping to lift the reliance on centralised control – when we allow it. Wikipedia and Ushahidi are just a few prime examples of decentralised information projects which have proved incredibly successful, in part because founders did not dictate what the ultimate result would be.
So often, I find that users of routing and scheduling software battle to achieve the results they want – not because the software is failing them, but because they refuse to accept the solution the software is presenting is the most apt. What appears logical when viewed as a group is not always in the best interests of the individual, or our individual opinions.
Take what American Air Liquide, a gas company situated in Texas, did when confronted with the growing problem of purchasing gas cheaply and efficiently delivering to over 6 000 sites. Using an algorithm to simulate ants foraging, the program would send out a billions of virtual trucks along their existing routes to determine the best paths for their drivers based on predicted sales volumes, production at various sites, weather conditions and price fluctuations. In effect, a program was dictating their entire logistic operations day-to-day.
Where drivers were going to the closest gas-producing plant to their customer and delivering from there, the fleet would now often travel much further to pick up gas from the lowest-cost gas producer – something that at the individual level of the drivers seemed counter-intuitive but, as a “colony”, saved them an estimated US$20 million annually.
The beauty of this solution was that it required very little terms of physical changes to the fleet makeup, organisation or infrastructure, which are standard responses to logistical challenges.
Air Liquide only found that they were holding onto gold by letting go.As usual, if you have any comments or questions please contact me at rick.de.klerk@opsi.co.za.
