Tutorials

These tutorials walk through complete end-to-end workflows for common tasks.

Deployment Workflows

• G1 Whole-Body Tracker: From Data to Real Robot - G1 whole-body tracker: from data to real robot (full pipeline)

Training Workflows

• SMPL Training on AMASS - Train SMPL humanoid on AMASS motion data

• Retargeting SMPL motions with PyRoki - Retarget AMASS/SMPL motions to robots

• Scene Workflow (Vaulting) - Scene interaction with vaulting motions

• Domain Randomization & Sim2Sim - Sim2sim transfer techniques

Robot & Environment Setup

• Adding a Custom Robot - Add your own robot morphology

Code Tutorials

For learning ProtoMotions internals from the ground up, see Code Tutorials (Progressive Series) - 8 progressive Python scripts in examples/tutorial/ that teach core concepts like simulators, terrain, robots, scenes, and environments.

Quick Reference

Workflow	Description	Prerequisites
G1 Whole-Body Tracker: From Data to Real Robot	Full pipeline: data to real robot (G1 tracker)	Pre-trained model or BONES-SEED data
SMPL Training on AMASS	Train SMPL on AMASS motions	AMASS data prepared
Retargeting SMPL motions with PyRoki	Retarget AMASS/SMPL to robots	Packaged AMASS .pt
Scene Workflow (Vaulting)	Full scene interaction workflow	Motion + scene data
Domain Randomization & Sim2Sim	Sim2sim transfer	Trained policy
Adding a Custom Robot	Add new robot morphology	MJCF file

Coming Soon

• MJCF to USD Conversion - Convert MuJoCo assets to USD for IsaacLab

• USD GS Backgrounds - Load Gaussian Splatting scenes in IsaacSim 5.0+

• Pure RL Training - Train without motion data (steering, locomotion)

• MaskedMimic / AMP / ASE - Generative policy training