Friday, August 28, 2009

Lobsters teach robots magnetic mapping trick

SPINY lobsters have become the unlikely inspiration for a robot with a unique sense of direction. Like the lobster, it uses a map of local variations in the Earth's magnetic field to find its way around - a method that could give domestic robots low-cost navigational capabilities.

In 2003, computer scientist Janne Haverinen read in Nature (vol 421, p 60) about the amazing direction-finding ability of the Caribbean spiny lobster Panulirus argus. A team from the University of North Carolina, Chapel Hill, had moved the critters up to 37 kilometres from where they were caught and deprived them of orientational cues, but found they always set off in the right direction home. They concluded P. argus must navigate with an inbuilt map of local anomalies in the Earth's magnetic field.

"My first inspiration came from birds, ants and bees," says Haverinen. "But the spiny lobster clinched it for me."

The findings set Haverinen, who works in the intelligent systems lab at the University of Oulu, Finland, wondering if he could draw magnetic maps of buildings for domestic and factory robots. It is well known that compasses are sent haywire by the metal in buildings - plumbing, electrical wiring and the steel rods in reinforced concrete, for instance - and cannot find magnetic north. Haverinen's idea was that these distortions of the Earth's magnetic field might create a distinctive magnetic topography.

"So we decided to try to use this 'magnetic landscape' - the array of disturbances - that was upsetting the compass as a map for a robot," says Haverinen.

The team used a magnetometer to scan the magnetic field strength close to the floor in their lab (see picture) and in a 180-metre corridor in a local hospital. They then stored the field variations in the memory of a small wheeled robot and mounted a magnetometer on a rod projecting in front of it to prevent interference from its motors.

The robot was able to work out where it was and to drive along the corridor without a vision system. What's more, the magnetic map stayed true a year after the first mapping was done, Haverinen reports in Robotics and Autonomous Systems (DOI: 10.1016/j.robot.2009.07.018).

"So there just might be enough stable information for robots to work out where they are in the ambient magnetic field," he says. That would obviate the need for expensive "indoor GPS" systems in which triangulation between fixed radio beacons in a building tells robots their position.

"Reliance on any one guidance method is not a great idea in case it fails," warns Chris Melhuish, director of the Bristol Robotics Laboratory in the UK. "But you could use a system like this, if it's proven to work, to boost your confidence in a robot by using it in conjunction with, say, vision-based navigation."

No comments:

Post a Comment