AI Robot Conquers Classic Maze Puzzle Game – Watch Now!

Robots have consistently outplayed humans in games like chess, poker, and Go. But it’s much harder for them to beat humans in games requiring physical dexterity. However, a group of researchers at Switzerland’s ETH Zurich have developed a new robotic system known as CyberRunner, that has managed to learn how to play the classic children’s puzzle game Labyrinth faster than any human.

Labyrinth generally consists of a box topped with a flat wooden plane that tilts across an x and y axis using external control knobs, and a maze featuring numerous gaps. The goal is to move a marble from start to finish without it falling into one of those holes. The game usually requires practice and patience to master, but CyberRunner reportedly mastered the game in just 5 hours and can now complete the maze in just under 14.5 seconds—over 6 percent faster than any existing human.

CyberRunner’s newfound maze expertise is credited to its real-time reinforcement learning and visual input from overhead cameras. It uses hours’ worth of trial-and-error Labyrinth runs to learn how to navigate the marble successfully along its route.

Furthermore, CyberRunner discovers shortcuts and faults in the maze design itself, creating its own Labyrinth cheat codes that sidestep the maze’s marked pathways. The project is completely open-source and aims to encourage other researchers to utilize and improve upon the program’s capabilities.

“Prior to CyberRunner, only organizations with large budgets and custom-made experimental infrastructure could perform research in this area,” said project collaborator and ETH Zurich professor Raffaello D’Andrea in a statement this week.