Humanoid Sample Pipelines#
The Humanoid Toolkit sample pipelines demonstrate end-to-end AI workflows optimized for Intel platforms. Review the validated hardware configuration before running a pipeline.
End-to-end reference pipelines demonstrating imitation learning, vision-based manipulation, SLAM, and LLM-guided task execution.
Find supported Humanoid runtime, middleware, sensor, and pipeline packages.
Sample Pipelines Overview#
Pipeline |
Domain |
Description |
|---|---|---|
Manipulation, AI |
Action Chunking with Transformers optimized with OpenVINO™ for ALOHA robots. |
|
Manipulation, AI |
Visuomotor diffusion policy for the Push-T manipulation task. |
|
Manipulation, Control |
ACT imitation learning with OCS2 MPC and MuJoCo simulation. |
|
SLAM, Perception |
Feature-based visual SLAM for monocular, stereo, and RGB-D cameras. |
|
Foundation Model, AI |
Bimanual manipulation foundation model running in MuJoCo and real ALOHA robots. |
|
VLA, Manipulation |
Vision-Language-Action pipeline with PaliGemma VLM and flow-matching policy. |
|
GenAI, Manipulation |
Voice and text-commanded robot control with Phi-4 and SAM/CLIP. |
|
Agentic AI, ROS 2 |
Agentic ROS framework demo with OpenClaw. |
|
LiDAR, SLAM |
Fast, robust LiDAR-inertial odometry package. |
|
LiDAR, Visual SLAM |
Fast LiDAR-inertial-visual odometry package. |
|
LiDAR, SLAM |
Robust LiDAR-inertial odometry via point-by-point integration. |
|
Control, Manipulation |
Whole-body control pipeline for humanoid robotics. |
|
AI Inference |
OpenVINO-accelerated GR00T foundation model deployment. |