humanoid-robotics

Curated embodied AI list: surveys, VLA models, datasets, simulators, humanoids, robot learning, and safety resources.

Raspberry pi based humanoid robot.

Predictive Style Matching (PSM), is method in which an offline predictor maps the robot’s lower-body state history and velocity commands to interpretable upper-body joint and gait targets that shape the rewards during training.

Low Level RL Controller for G1

This repository contains the python version of the source code for the experiments carried out for the On-line Object Detection and Instance Segmentation project.

Complete Physical AI & Humanoid Robotics textbook with 87 interactive lessons, AI chatbot, ROS 2, NVIDIA Isaac Sim, and hands-on projects. Built with Docusaurus.

Test code for LAC: MuJoCo sim2sim stack and trained policy checkpoint on the Unitree G1

Autonomous path planning system for the NAO humanoid robot using the A* algorithm, grid-based world modelling, Manhattan heuristics, obstacle avoidance and real-robot execution for ball-retrieval tasks. Includes simulation in Webots/RViz and real NAO execution.

General Motion Retargeting on the GPU!

Velocity tracking and sim-to-sim transfer for Unitree H1 humanoid locomotion using reinforcement learning.

MINI_DIO ist ein Forschungsprojekt zur MCM-basierten Innenfeldreaktion, emergenten Bedeutungsverdichtung und passiven Selbstorganisation in einem kleinen künstlichen Wahrnehmungssystem.

Official implementation of Humanoid-Motion-Diffusion, leveraging generative diffusion models to synthesize advanced motion controllers for humanoid robots and legged locomotion research.

Open-source 7-DOF 3D-printable humanoid robotic arms, Supports ROS2, vision, and autonomous control. Built for under $800.

Fuze — surprise-gated metaplastic consolidation for continual learning. No replay buffers, no task boundaries, no Fisher passes. Built to become the learning core of humanoid-robot AI.

A HRI game in which you guess objects with a robot.

Open-source robotics simulator for coding agents: HTTP/JSON + MCP control, Newton physics, wgpu rendering, ROS 2, and reproducible benchmarks.

Unitree H1 humanoid walking via DeepMimic-style imitation learning — MuJoCo MJX + JAX, 4096 parallel envs, 9000 FPS, full 19-DOF control against real human mocap data

Evaluating whole-body control architectures (WQP, HQP, Passivity, LQR) for humanoid balance in Mujoco simulation with learning-based residual compensation. MSc thesis @ University West 2026.

Real2Sim MVP for Unitree G1: human pose retargeting to MuJoCo with OpenClaw-compatible launchers.

Open-source motion compiler for validated, cross-vendor humanoid robot trajectories