Interactive Reinforcement Learning
Train a tiny dinosaur in a grid world. Place food, lava, and rocks, then watch Q-learning discover safe, efficient paths.
Step 1: Learn to Move
Watch the agent explore the grid with no goal yet.
Tip: You can still place objects manually. Apply Step resets the environment.
Click the grid to place items. Use Erase to remove. Start is where each episode resets.
The agent learns to reach food while avoiding dangerous or costly tiles.
With ε-greedy strategy, it sometimes tries new actions to avoid local optima.
When the world changes, the agent keeps learning rather than memorizing one path.