Unity-Technologies/ml-agents

Project overview

This project provides Unity and Python tooling for training intelligent agents in games and simulations using reinforcement learning, imitation learning, neuroevolution, or custom methods. It includes PyTorch-based trainers, a Python API for environment control, and cross-platform inference support.

Project type

AI Agent · Model Development · Evaluation & Observability

Deployment

Refer to project documentation

License

License pending

Best for

• Researchers and developers needing to train and evaluate intelligent agents within Unity games and simulations using reinforcement learning or imitation learning.

Key capabilities

• Enables games and simulations to act as environments for training intelligent agents.
• Provides PPO, SAC, MA-POCA, and self-play support for single-agent, cooperative multi-agent, and competitive multi-agent training scenarios.
• Supports learning from demonstrations through BC and GAIL algorithms.
• Allows users to add custom training algorithms and training components.
• Supports definable curricula for complex tasks and environment randomization for training robust agents.
• Includes more than 17 example Unity environments that users can study or adapt.
• Provides a Python API for controlling Unity environments and training agents with supported or user-defined methods.
• Uses the Inference Engine to provide native cross-platform support for trained-agent inference.

Limitations and risks

• The develop branch is under active development and may be unstable.
• The toolkit includes in-editor analytics associated with information passively collected by Unity. It is unclear whether collection can be disabled.

Getting started

• Setup difficulty and first success paths are not documented in the provided materials.

Evidence and sources

• GitHub project description: The Unity Machine Learning Agents Toolkit (ML-Agents) is an open-source project that enables games and simulations to serve as environments for training intelligent agents using d…
• README: We provide implementations (based on PyTorch) of state-of-the-art algorithms to enable game developers and hobbyists to easily train intelligent agents for 2D, 3D and VR/AR games.
• README: Researchers can also use the provided simple-to-use Python API to train Agents using reinforcement learning, imitation learning, neuroevolution, or any other methods.
• README: These trained agents can be used for multiple purposes, including controlling NPC behavior (in a variety of settings such as multi-agent and adversarial), automated testing of gam…
• README: - 17+ [example Unity environments](https://docs.unity3d.com/Packages/com.unity.ml-agents@latest/index.html?subfolder=/manual/Learning-Environment-Examples.html)

Reviewed at: 2026-07-16T17:39:28Z