Dial the number, let it ring. That’s usually all it takes to contact anyone else in the world in possession of a telephone. We take this simplicity for granted, but for some time towards the end of the 20th century, there was serious concern that the rapidly increasing demand on telephone networks was not going to be met.
In the early days of the telephone, calls were reserved for urgent communications only. But times have changed. According to Ofcom, the UK spent 103 billion minutes on the phone in 2012 using landlines alone. Managing this volume of call time presents a significant challenge requiring continually new solutions – and not always from obvious sources.
A phone network consists of many individual phone lines and intermediate ‘nodes’ – servers or bases that act as in-between points during the call. Every time you make a phone call, a path, or ‘route’, between your phone and the phone belonging to the person you’re calling has to be secured. This route is constructed by connecting varying numbers of nodes between the two phones. In the early 20th century, these routes were supervised by human operators, but by the second half of the century, the process had become almost entirely automated, and a central system made all decisions about which routes were used.
By 1990, the UK phone network managed over 30 million phone numbers, and telecommunications companies such as British Telecom were facing trouble. The problem was that the quality of connections within the phone network could change, and the central system was having difficulty keeping up. Just as a road accident could make you drastically reconsider the best route between two cities, a broken or extremely busy connection between nodes in the phone network could significantly alter the best call route between two phone lines. The only way to adapt to these changes in the network was by running time-consuming scans and updating information about all the possible routes between nodes, causing delays and putting the system under strain. In 1994, BT made the sobering estimate that a traditional approach to improving the national phone network – for example, by repairing broken lines – could cost over 25 billion pounds and take a decade to complete. There had to be another way.
Meanwhile, several hundred miles away at the Polytechnic University of Milan, an Italian PhD student called Marco Dorigo had begun watching trails of ants. He had noticed that trails nearly always took the most direct route between items of food and the nest. What’s more, the ants made these trails in a remarkably simple way.
When they find food, ants lay down a chemical called a pheromone which acts as a guide for other ants. Those ants travelling by more direct routes would reach the food sooner, and so the amount of pheromone being laid on their route would be increased more quickly than the pheromone on more indirect (and therefore slower) routes. After only a few ant journeys, a colony could find the shortest route between the food and the nest.
Dorigo made the connection. In his PhD thesis he demonstrated how simulated ants on a computer could use virtual pheromone trails and quickly find the shortest route between two points through a maze – or through a network. Importantly, each virtual ant followed only a few very simple rules, and there was no power-guzzling central system organizing the process. Instead, everything was recorded in the pheromone trail.
This new, decentralized approach was Nature’s answer to the very problem that had been plaguing telecommunications. In 1996, a team of researchers lead by Dr Ruud Schoonderwoed at Hewlett-Packard’s laboratories applied this solution to telephone networks. Again using virtual ants laying simulated pheromones between nodes in the network, they created a system that could quickly find the least congested available route between two points. At each node, the highest concentration of virtual pheromone indicated the best path to the next node, so that no central system was required to keep track of all the possible routes across the network. Rather than mend broken or congested connections between nodes, the virtual ants allowed the network to bypass these connections altogether and find alternatives.
The telecommunication world was impressed. In the late 90s, British Telecom began adopting a similar system to their networks, and the French equivalent, France Telecom soon followed suit. The so-called ‘Ant Colony Optimization algorithms’ allowed the network to manage itself.
The idea has since inspired many solutions in other networks, including the internet and flight networks connecting airports. As Dorigo noted in 2004, we might do well to heed the proverb: “Go to the ant, thou sluggard; consider her ways and be wise.”