21 January 2000

Avalanche rescue

by Kate Melville

Skiing and snowboarding might be great fun, but getting caught in an avalanche can really ruin your holiday and will quite possibly end up in your final method of transport home being a pine box.

To be blunt if you are not found quickly (generally within thirty minutes) at best you can look forward to dying of, frostbite, hypothermia, suffocation, or a combination of all three.

Is the danger real? Well the growth of 'New School' skiing and snowboarding has literally seen an 'explosion' in the number of people seeking backcountry and off-piste experiences (33 deaths in the USA in 1999). This change has unfortunately been mirrored by record numbers of avalanche deaths in both Europe and North America. While many recent deaths occurred in villages and locations where local were well aware of the dangers, many of the people venturing to the mountains are only vaguely (if at all) aware of avalanche risk and how to mitigate them.

Many backcountry sports people already carry radio beacons for avalanche safety. The current rescue technique is to both exhaustively search every inch of the slide path and exit line or for the searcher to make a right-angle turn when signal strength from the radio beacon decreases. This is a basic system that has problems with things like buried obstacles masking signal strength, and the fact that the transmitter's physical orientation is unknown.

So now science has help come up with a new if somewhat unconventional way to help.

According to Rush Robinett the US Department of Energy's Sandia National Laboratories finding a buried skier is akin to locating the source of a nuclear, biological or chemical warfare attack. The military version uses software where group intelligence is used for a series of small robots to be used to rapidly locate a source of contagion.

The US Department of Defense version of the algorithm allowed cockroach-sized robots to communicate with each other via miniature radio transmitters and to then locate a target far more quickly than using conventional means. Known as 'swarming', each robot constantly swaps data with other robots about its position and of the signal strength at that position relative to the sought-for source. This continuous data update from multiple sources allows each 'swarm' member to continually refine the direction of its search, rather than relying on intuition or a central intelligence telling the searchers what to do.

Robinett says that traditional search communications has been very basic with, "Yelling and pointing are probably not the most efficient ways to conduct a rapid search". The new system mimics the military version by using a group of people carrying a GPS, linked to a mini-computer and radio to help find people buried under an avalanche.

The system that uses a "swarm" algorithm called "Distributed Optimization" found avalanche victims up to four times more rapidly (in simulations) than any existing system.

Now what works in the controlled environment of a research lab may not translate well in the field? Robinett acknowledged this and said that the initial results were, " with conditions as simple as possible". Yet in a surprise result the 'swarm' algorithm worked even better in situations, where things like trees, rocks, and snow depth created complications.

In the near future Robinett envisions that, "we can solve a whole class of similar problems with the same algorithms. By carrying lightweight radios, GPS positioning devices, and pocket computers programmed with robotic search algorithms, rescuers would be told [via computer screen], 'South three steps for most efficient path search,' or some such thing."

Hopefully this research translates into new search and rescue techniques soon, although one might argue that dying in an avalanche is simply part of 'natural selection' in human evolution?