Games & virtual worlds · PROJECT NOTES
Unity ML-Agents
A Unity SDK and Python interface that turn game and simulation scenes into interactive environments for reinforcement learning, imitation learning and multi-agent research.
What can you do with it?
Train and compare policies in instrumented Unity scenes using vector or visual observations, continuous or discrete actions, demonstration data and cooperative or competitive agents.
What should you know before using it?
An environment toolkit, not a fixed benchmark or game-generation system. Release 23 documents Unity 6000.0+, C# package 4.0.0 and Python packages 1.1.0 with Python 3.10.1–3.10.12; source and package versions must be pinned separately. The reviewed OpenEnv wrapper uses only the first agent and does not support multi-agent environments; its x86_64 Docker binaries do not run on Apple Silicon.
What is verified here?
Creator: Unity Technologies and ML-Agents contributors
Runtime: Authoring or modifying scenes needs the Unity Editor and matching ML-Agents package; Python controls the Editor or a platform-specific executable. The pinned OpenEnv wrapper can download example binaries and run a local server; graphics mode needs a display. Desktop tags describe documented environment runtimes, not coverage of every Unity export platform.
License note: ML-Agents toolkit code is Apache-2.0, with separate third-party notices. Unity Editor software has its own terms; the toolkit license does not replace engine, dependency or external-asset terms. Upstream license ↗
This entry was reviewed against original project material. We have not independently run the software or reproduced its results. AI-development claims are author-reported; check upstream for current versions, setup requirements, and licenses.
- Resource type
- Environment framework
- Provenance
- Development provenance not established
- Reviewed
- Platforms
- Windows, macOS, Linux