ml-agents

Overview of ml-agents

What it is

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. It provides implementations based on PyTorch of state-of-the-art algorithms for deep reinforcement learning and imitation learning. The toolkit is designed for game developers and researchers to train agents for 2D, 3D, and VR/AR games.

Key features

• 17+ example Unity environments - Support for single-agent, multi-agent cooperative, and multi-agent competitive scenarios via PPO, SAC, MA-POCA, and self-play - Imitation learning through BC and GAIL algorithms - Flexible Unity SDK that can be integrated into custom Unity scenes - Curriculum learning and environment randomization for robust training - On Demand Decision Making and multiple concurrent Unity environment instances - Cross-platform support via the Inference Engine and Python API with gym/PettingZoo wrappers

Use cases

Trained agents can be used for controlling NPC behavior in multi-agent and adversarial settings, automated testing of game builds, and evaluating different game design decisions pre-release. Researchers can use the Python API to train agents with reinforcement learning, imitation learning, neuroevolution, or custom methods.

Getting started

See the repository README for installation instructions and links to the Unity Package documentation.

License and pricing

The license is not stated in the metadata (NOASSERTION). The project is described as open-source and is free to use.

Based on the project's public README and repository metadata.