This robot can beat humans in the wooden maze game, which is a huge achievement

Artificial intelligence (AI) now knows how to beat humans in many games that rely on logic or strategy, such as chess, Go, or video games. However, it is different when it comes to taking the laws of physics into account. However, here is CyberRunner, an AI-guided robot system capable of mastering the ball maze game. It is the work of researchers from the Swiss Federal Institute of Technology in Zurich.

The maze game requires a certain amount of dexterity to guide the ball along the path without falling into one of the many holes. To do this, the player must use both handles to tilt the maze back and forth, or from left to right.

New record time for this robot

The researchers relied on recent advances in model-based reinforcement learning. The AI ​​learns as you play thanks to the camera placed above the maze, thus gathering experience. It took him just six hours to learn how to guide the ball across the entire track in record time, beating the human best time by 6%. The researchers also had to intervene to force the AI ​​to follow the path because it was “cheating” after discovering the shortcuts.

Comparison between human players and CyberRunner in the maze game. In English, enable automatic translation of subtitles. © ETH Zurich

” Now, for less than $200, anyone can embark on cutting-edge AI research in the physical worldRaffaello D’Andrea, one of the study’s authors, noted. Additionally, once there are thousands of CyberRunners in the real world, it will be possible to conduct large-scale experiments, where learning occurs in parallel, on a global scale. The researchers plan to release the source code and hardware specifications for CyberRunner under a free license On their website.