CyberRunner

Background:
CyberRunner is an innovative project centered around an AI robot designed to master the labyrinth marble game. This game, a test of physical skill and strategy, involves navigating a marble from a start point to an end point on a board laced with pitfalls. Players control the marble's movement indirectly using two knobs that alter the board's orientation. Despite its apparent simplicity, the labyrinth game demands fine motor skills and spatial reasoning, skills that humans typically develop with extensive practice.

Research and Development:
CyberRunner harnesses the latest advancements in model-based reinforcement learning. This involves the robot learning through experience, much like a human player. It captures observations and evaluates its performance through a camera, effectively serving as its "eyes" on the labyrinth. This continual acquisition of experience allows the robot to refine its understanding of the game mechanics, identifying more effective strategies and behaviors. Unlike traditional learning methods, CyberRunner's learning process is concurrent with gameplay, meaning the robot continuously evolves and improves its skills with each run, without the need for pauses or downtime.