Skip to content

README.md

Latest commit

 

History

History
37 lines (26 loc) · 1.28 KB

README.md

File metadata and controls

37 lines (26 loc) · 1.28 KB
Raw

"# AI-Sim

A browser-based 2D world simulation demonstrating client-side AI inference and training using TensorFlow.js. Published as a GitHub Page.

What it does

Two species of creatures live in a tile-based world rendered on Canvas 2D:

  • 🌿 Herbivores — seek food, flock with each other, flee predators
  • 🦊 Predators — hunt herbivores, hold territory

Each species shares a CNN brain (TensorFlow.js) trained in-browser via Deep Q-Network (DQN). Creatures perceive a 7×7 tile grid around them and learn which actions (move up/down/left/right/stay) maximise their rewards.

Player controls

Use the sidebar toolbar to:

  • 🌱 Paint food sources
  • 🟩🟦🟥 Paint terrain (grass / water / danger zones)
  • Spawn or clear creature populations
  • Tune the replay buffer size (1 = online learning, 2000 = stable mini-batch RL)
  • Adjust simulation speed and exploration rate (epsilon)

Tech stack

  • TensorFlow.js — in-browser ML
  • Canvas 2D — rendering
  • Vite — build & dev server
  • GitHub Actions — CI/CD to GitHub Pages

Development

npm install
npm run dev       # local dev server
npm run build     # production build → dist/

Push to main to deploy automatically to GitHub Pages."