Q-LAB: Gridworld AI Simulator

A high-performance, zero-dependency visualization of Reinforcement Learning running entirely in the browser.

Watch an AI agent learn to navigate complex mazes using the Q-Learning algorithm. You can customize the environment layout, adjust hyperparameters in real time, or toggle Turbo Mode to train the agent at maximum computational speed.

Live Demo

• Real-time Tuning: Adjust learning rates, exploration rates (epsilon), and discount factors (gamma) dynamically via sliders.
• Turbo Mode: Speed up execution past 85% to disable rendering and train the agent at maximum speed.
• Interactive Map Editor: Draw custom grids with walls, mud (slow zones), ice (slippery zones), and portals for teleportation.
• Visual Overlays: Toggle visualization overlays including Policy Arrows, Heatmaps, and the optimal path layout.

This project has no external build steps or dependency installation required.

1. Download the index.html file.
2. Open index.html directly in any modern web browser.
3. Use the control panel to customize parameters and start training.

• Single-File Architecture: HTML, CSS, and JavaScript are bundled in a single standalone file.
• HTML5 Canvas: Used for high-frequency grid and path rendering.
• Chart.js: Handles live plotting of scores and training metrics.

MIT. See LICENSE.