Gym environment for AI experimentation/research in Minecraft
Top 93.6% on SourcePulse
This project provides a set of OpenAI Gym environments for interacting with Minecraft, leveraging Microsoft's Malmo platform. It's designed for AI researchers and developers experimenting with reinforcement learning agents in a 3D block-based world, offering various pre-configured Minecraft missions for training and evaluation.
How It Works
The environment bundle integrates with the Malmo platform, which acts as a bridge between Minecraft and AI agents. It exposes standard Gym step
, reset
, and render
methods, allowing agents to perceive the Minecraft world state, take actions, and receive rewards. The environments are based on Malmo's missions, with added PrioritiseOffscreenRendering
to potentially speed up training by reducing visual overhead.
Quick Start & Requirements
pip install gym
.minecraft_py
: git clone https://github.com/tambetm/minecraft-py.git && cd minecraft-py && python setup.py install
.gym-minecraft
: git clone https://github.com/tambetm/gym-minecraft.git && cd gym-minecraft && python setup.py install
.pygame
for rendering.Highlighted Details
Maintenance & Community
MalmoEnv
instead.Licensing & Compatibility
Limitations & Caveats
The project is marked as outdated, with a recommendation to use MalmoEnv
. This suggests potential lack of maintenance, compatibility issues with newer Gym versions, or unaddressed bugs. The setup process involves downloading and compiling Minecraft, which can be resource-intensive and time-consuming.
4 years ago
1 day