ANTS from Central America are helping engineering design to lose its
reputation as a black art. French researchers say the ability of
Pachycondyla apicalis ants to hunt in parallel could dramatically improve
the way computers search for the solutions to complex problems such as aircraft
and turbine design.
Unlike most ants, P. apicalis forages by dispatching lone hunters to
scour an area and report back to the colony. By using random patterns and
concentrating on areas near the nest which have previously proved fruitful, the
colony can carry out a large number of searches simultaneously, making it a
highly effective searching system.
鈥淲hat we have done is make a [software] model of this foraging strategy and
apply it to benchmark problems,鈥� says Gilles Venturini, a computer scientist at
the University of Tours. He and his team have started with so-called 鈥渘umerical
optimisation鈥� problems in which the best solution must take a number of
different variables into account. For example, what shape should a turbine be,
given its size, speed, the fluid it must operate in and the required rate of
flow?
Advertisement
Traditionally, such problems have been solved as much by trial and error as
science because they are so complex. However, Venturini and his colleagues found
that even a crude version of their ant algorithm worked as well as existing
techniques, or better, and they hope to improve it with further work.
鈥淎 very great advantage is that the algorithm can be implemented easily in
parallel,鈥� says Venturini. Parallelism is important because each 鈥渁nt鈥� can work
on a different computer, searching part of the numerical landscape for the
lowest point, for example. The colony moves when the surrounding area has been
explored. The team now plans to apply the idea to Web searches.