Soldiers carry a 75 kilogram (165 pound) dummy onto a practice field. For the purposes of the M-ELROB competition, that is an injured soldier who needs saving. The robots have to get the dummy off the field, which in a real situation could be littered with explosive mines.

A robot makes its way through waist-high grass. It has the GPS-coordinates of the dummy's approximate position. But the person controlling the robot can hardly see what's going on: they are far away and can only monitor the scene via video.

The robot finds the dummy. But how will it get the dummy - injured soldier - to safety?

The robot picks up the dummy-soldier by its vest. It lifts the dummy up as high as possible. This is important because you shouldn't drag an injured person across a mined field.

The robot falls over and takes the "injured soldier" with it. Both are down, lying in the grass. The only solution is for the robot to let go of the dummy and try to get it up again. But it's a tough ask, and the robot fails.

Meet Marek, a remote-controlled robot from the military academy in Warsaw. Marek has a grabber with a camera and a large excavator shovel. It can pick up tree trunks - so humans are no problem at all.

The lower shovel digs into the ground without touching the dummy and lifts it right up.

Marek saves the "injured soldier" in seconds. The dummy lies safely in the shovel, cushioned by straw and soil. In a real situation, the robot would now take the soldier to a safe area where his or her injuries can be treated.

But if there's no room for a huge excavator, this nifty little number from the Fraunhofer Institute for Communication, Information Processing and Ergonomics (FKIE) will do the trick. It can even climb stairs and get around tight corners.

A robot this small couldn't carry an injured person. It uses a tow rope instead, which attaches with a hook.

But without a belt, the robot wouldn't stand a chance. Fortunately, the dummy is wearing one. So the robot has a number of options to place the hook. A sailor's life vest, for instance, would also be very practical for this purpose.

The human "pilots" control the robots from a dark tent far away from the action. They can't see the robots and have to rely fully on cameras and other sensors attached to the remote rescuers.