There are a whole heap of reasons why the Army's new squad of killer drones will be remote-controlled by blood-and-guts soldiers, instead of roaming around Iraq on their own.

One is that unmanned ground vehicles, or UGVs, still aren't particularly good at navigating around uncertain terrain.

Just look at what happened during last Spring's Pentagon-funded robot rally through the Mojave Desert. No robo-car made it more than a few miles without getting stuck in a ditch somewhere.

Everybody at his roboticist brother has a solution for fixing this. One idea over at the Army's Tank-Automotive and Armaments Command, or TACOM, is to load the UGV up with satellite imagery, to let the bot know what's ahead.

Robots today often get elevation data – how high the obstacles are in its way – but the consistency of the ground is often unknown. But by snapping pictures in the visible, infrared, and other ranges, a satellite can cobble together a pretty decent picture of the land below: what's mud, what's solid earth, and what's a pile of rocks.

With that information in hand, a soldier can plot out a path for the UGV that gives the robot a better chance of successfully making it through the trip. The variations are small, but crucial, TACOM engineer Phil Frederich said. In testing at Ft. Knox about two months ago, an unmanned Stryker, driving on roads, often moved from the middle by a foot or two. It wasn't much. But it helped the robot make it through the course in one piece.