After a few months of tests in the lab, the team plans to take the new prototype to Sedan Crater, in Nevada. This crater, formed by an underground nuclear-test blast, has steep slopes of up to 40 degrees composed of loose sand. The steepest inclines in last month's Meteor Crater tests were about 23 degrees and composed of rocky ground. Unlike wheeled rovers, the tetrahedral structure should be able to handle the steeper incline without slowing down.

The team has designed versions for both lunar and Mars exploration in hopes that they will be able to crawl into narrow cracks and around, over, and under obstacles that a wheeled rover can't negotiate. More-advanced versions could even "chimney" up a space between vertical cliff faces. (This technique was developed by rock climbers to enable them to ascend a narrow vertical space: they lean against one wall with their legs against the other and gradually inch upward.) And since there's inherently no risk of the device falling over and it is unlikely to get stuck, the vehicles could take more chances in exploring the difficult nooks and crannies that might be of great interest geologically and biologically.

Working in conjunction with conventional wheeled rovers, the two kinds of vehicles could form an effective team: the wheeled vehicles would be able to cross long flat distances and serve as a base, while the "tets" could scurry around as scouts, looking for the most interesting places and retrieving samples.

The devices could also find applications on Earth. Wheeled robots are now sometimes used for hazardous tasks such as finding land mines and exploring dangerous chemical-pollution sites and volcanic calderas. But the tetrahedra could do so in places the wheeled versions could never get to.

David Wettergreen, a robotics engineer at Carnegie Mellon University, says the Goddard design could be a useful adjunct to other rover designs. "You often want to specialize your mode of mobility for particular applications," he said, and it is therefore useful to have different mechanisms available. But he adds that the computational challenges could be daunting: even working out different gaits for robots with four legs is a difficult task. "This could have huge potential, but it will also be a big challenge to figure out how to use that potential."