The 15-pound robot uses 3-D cameras on its head and touch and force sensors on its claws to climb. The real trick, however, is in how it maintains its balance. As it climbs, the robot uses a neural network to teach itself to balance as it transitions from vertical to horizontal movement and moves its legs up the rungs.
Teaching it to climb ladders of any dimensions without pre-programming it is the next step for the researchers, as the robot was specifically programmed ahead of time to maneuver the ladder in the video below.