ProtoMotions Documentation

ProtoMotions is a GPU-accelerated simulation and learning framework for training physically simulated digital humans and humanoid robots. Our mission is to provide a fast prototyping platform for various simulated humanoid learning tasks and environments, bridging efforts across physics-based animation, digital humans, and humanoid robotics.

Note

This project will download and install additional third-party open source software projects. Review the license terms of these open source projects before use.

Key Features

  • GPU-Accelerated: Fast, scalable simulation for reinforcement learning workflows with NVIDIA Newton (beta), IsaacGym, IsaacLab, and Genesis backends

  • Modular Design: Multiple simulation backends, robot morphologies, RL environments, and algorithms with built-in support. Add your own robot, task, or algorithm with ease

  • Rich Toolkit: Built-in procedural terrain generation, motion retargeting (PyRoki-based), scene and object spawning. All scalable to large training runs

  • State-of-the-Art Algorithms: MaskedMimic, AMP, ASE, PPO implementations

  • Multiple Robots: SMPL, SMPL-X, Unitree G1, H1, and custom morphologies

  • Open Source: Permissively licensed under Apache-2.0

High-Level Architecture

ProtoMotions Architecture