Using reinforcement learning to design an intelligent AI for Minecraft. The model uses image analysis to "see" it's world and optimizes play in a closed system
You must be running Windows to run this AI (We require keybindings,screen capturing, and mouse control through the win32api)
Create a conda environment using
conda create --name ENV_NAME python=3.4
Install from the list of requirements.
pip install -r requirements.txt
To run the AI, do the following. We assume that you have already started your conda environment using
source activate ENV_NAME
- First, we must configure Minecraft to the correct settings. As of the moment, you must set the following options:
- Resolution: 1366x768
- Block Distance: less than 10
- No Light Shading
- Windowed Mode
-
Start Minecraft
-
Run the following command
python -i start.py
main()
A prompt will appear to switch to your minecraft window: do so, and watch the AI run
Allow computer to control game inputsGet display- Machine learning algorithm for detecting block