Grid: 6×6 | Actions: ↑ ↓ ← →
⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.
This simulation showcases a robot learning through reinforcement learning. A wheeled robot starts in the corner of a 6×6 grid world scattered with walls, a goal tile and a hazard pit, and repeatedly attempts to cross it. Every move updates a tabular Q-learning table from the reward or penalty it receives, exploration gradually gives way to exploitation as epsilon decays, and the floor lights up with a live heatmap of the learned values plus arrows showing the current best-known policy — so you can watch the robot's behaviour visibly improve, episode after episode, purely from trial, error, and reward.