US scientists are planning the smallest space invaders of all intelligent robot probes weighing no more than half an ounce could one day be scurrying about the solar system, boldly going where no satellite could go before.

Until now, engineers have designed satellites to withstand the harshest environments, but Mr Kurt Moore and Mr Mark Tilden of the Los Alamos National Laboratory in New Mexico have another idea altogether: they will give their probes a rudimentary little brain and let them work out for themselves how best to survive in a lethal planetary radiation belt.

Every satellite sent up is a multi-million dollar gamble. Nearly half of all failures are because of loss of radio contact with mission control back on earth, usually because the radiation blitzes dainty microprocessors.

But the Los Alamos team has another approach. "We are working on satellites that have no microprocessors or fixed algorithmic behaviours," said Mr Moore at a meeting of the American Geophysical Union in San Francisco. "These satellites are survivors designed from the bottom up and domesticated by their sensors and control payloads into performing high reliability tasks."

The control systems are modelled on the simplest "twitches" of animal neurons. The team has a name for them: biomorphs. Mr Tilden has already built 200 types of biomorphs or mechanical insects based on a control loop which sends a message from one leg to another to keep the robot moving. The system is almost impervious to electrical and mechanical faults. The little robots have been tested in a US military firing range full of unexploded ordnance: if a leg gets blown off a quadruped robot, it simply works out how to carry on with three, or two, or even one leg.