AI Path Finding Game

Watch the AI learn in real-time through trial and error

⚠️ About the AI Training

This AI runs entirely in your browser using TensorFlow.js, which means:

  • Training progress is not saved between sessions
  • Learning speed depends on your device's processing power
  • The AI starts from scratch each time, with no prior knowledge

Please Note: The AI learns through exploration and experimentation. Initially, it will make random moves to understand the environment. This means:

  • Early movements may appear random or inefficient
  • Learning progress becomes visible after several episodes
  • The AI gradually develops better strategies through trial and error
  • Patience is key - meaningful improvements typically show after 50+ episodes

Game Elements

Player: AI-controlled agent
Food: +20 energy
Poison: -10 energy
Bombs: -30 energy
Target: Goal position

Quick Guide

Training Parameters

  • Epsilon (ε): Controls exploration vs. exploitation
  • Learning Rate: Speed of knowledge acquisition
  • Gamma: Future reward importance

Objective

The AI must reach the target while maintaining maximum energy. It learns to:

  • Collect food for energy
  • Avoid poison and bombs
  • Find optimal path to target

Game Environment

Energy: 100

Training Dashboard

Learning Parameters

1.00
0.0010
0.99

Training Stats

Episodes
0
Average Reward
0.0
Food Collected
0
Current ε
1.000

Performance Charts

Reward History

Loss History