LemmoLab: すべてのプロダクトが 50% 割引

Unray Plugin for RL

Training Tool for Multiagent Scenarios with Reinforcement Learning in Unreal Engine

  • サポートされたプラットフォーム
  • サポートされたエンジンバージョン
    5.3
  • ダウンロードのタイプ
    エンジン プラグイン
    この製品には、コード プラグインが含まれており、ビルド済みのバイナリと Unreal Engine に統合される全ソースコードが完備されています。任意のエンジン バージョンにインストールし、プロジェクト毎に有効化することが可能です。

Discover how Unray can power your game and simulation development with reinforcement learning in Unreal Engine:


1. Interactive Game Development: Use Unray to create complex game environments with multiple agents that learn and adapt as they play.

2. Realistic Environment Simulations: Create realistic simulations to train agents in environments that mimic real-world situations.

3. Research in Artificial Intelligence: Employ Unray as a research platform to experiment with different reinforcement learning algorithms in multi-agent environments.


Features:


- Uses powerful RLlib technology for effective training.

- Leverages the ability to parallelize training using Ray technology.

- Supports a variety of algorithms, including PPO, QMIX, DQN, in addition to those built into the RLLib library.

- Facilitates the creation of multi-agent environments.


Demo Video: https://youtu.be/6lu0gTPYFzY

テクニカルノート

Features: (Please include a full, comprehensive list of the features of the product)

  •  Train single and multiagents envs with Reinforcement Learning
  •  Parallel Training
  • Create and configure agents for RL training
  • Create envs for RL training

Code Modules:

  •  Name: Unray. Type: Runtime.

Number of Blueprints: 9

Number of C++ Classes: 1

Network Replicated: No

Supported Development Platforms: Windows

Documentation: https://github.com/Nullspace-Colombia/unray-bridge/tree/master

Important/Additional Notes: Unray plugin is complimented with a Python API counterpart, which makes use of RLLib, so it is necessary to install and develop the training from a Python IDE.