Grid World AI

This is a 2D demonstration of a grid with a goal coordinate (red square) and an AI that tries to reach the goal (blue square). The program first asks for input regarding the size of the grid or the grid slices. The default number of grid slices is 5.

The algorithm used for the AI is called Q-learning. It is a reinforcement learning algorithm and doesn't require a model. It learns through trial and error. The training is a long process and takes even longer if the grid is large. The number of training episodes is 10000.

Warning

If you get an IndexError, please use a different grid slice.

Getting Started

Dependencies

Python
Poetry

Executing program

git clone https://github.com/rakin406/grid-world-ai.git
cd grid-world-ai
poetry install
poetry run python grid_world_ai/main.py

Author

Rakin Rahman

License

This project is licensed under the MIT License - see the LICENSE file for details