3. Introduction to engine2learn

3.1. What is engine2learn?

Engine2Learn consists of two components:

1. An UnrealEngine 4 (UE4) plugin (C++), allowing game developers to connect their game to a state-of-the-art machine learning (ML) environment in order to create better game AIs and smarter NPCs.
2. A python3 machine learning (ML) library with a focus on reinforcement learning (RL) and deep learning (DL) methods. Engine2Learn also features an automatic deploy/build/cooking-procedure for running games headlessly on Linux clusters and/or virtual machines (e.g. Vagrant). This way, games can be run in parallel on hundreds or thousands of nodes to accelerate learning and to maximize exploration speed for RL-algorithms.
3. Both engine2learn components (UE4 plugin and python lib) connect and talk to each other via TCP sockets and a simple msgpack-based protocol. Information flows from the game to the ML-environment in the form of “observations” (e.g. camera rendering results, positions of actors, etc..) and from the ML-environment to the game in the form of action instructions (e.g. “moveRight”, “Fire”, “Jump”, etc..).

3.2. Main Features

3.2.1. Simple, yet Powerful

Engine2Learn is very lightweight. It comes as a pip-installable python module (pip install engine2learn) on the client machine-learning side as well as a UE4 plugin for the game environment side.