Blog

Latest Posts

FluxVLA Engine: Train, Eval, Deploy VLA

A practical FluxVLA Engine guide from LeRobot datasets to training, evaluation, inference, and open-source VLA manipulation deployment.

fluxvlalerobotvla

9/18/202613 min read

Whole-body VLA

TacPAC: Tactile Action-Chunk Correction

9/16/2026

Whole-body VLA

Show-Harness: VLMs Play Robots with GUMI

9/14/2026

Browse Series

cuRobo on Jetson AGX Orin and Unitree G1

· 10 parts

Part 1/10

Ai Manipulation Agents

· 5 parts

Part 1/5

Why Multi-Agent Pipelines Beat Single VLA Models | AI Manipulation Agents #1

11 min read

Part 2/5

Perception Agent: Florence-v2 + AnyGrasp | AI Manipulation Agents #2

14 min read

Part 3/5

ALRM: Code-as-Policy vs Tool-as-Policy in ReAct | AI Manipulation Agents #3

11 min read

Part 4/5

Agentic Robot: SAP Protocol & Temporal Verifier

13 min read

Part 5/5

Sim-to-Real Deploy: Moving the SAP Pipeline from LIBERO to Physical Robots | AI Manipulation Agents #5

15 min read

Vima Humanoid Manip

· 5 parts

Part 1/5

VIMA Architecture: Cross-Attention Transformer Explained

10 min read

Part 2/5

VimaBench: 17 Tasks and 4-Level Generalization Protocol

15 min read

Part 3/5

Object Tokenizer: From Raw Pixels to Object Tokens with Mask R-CNN + ViT

15 min read

Part 4/5

Dataset 650K: Collecting Large-Scale Multi-Task Manipulation Data for VIMA

12 min read

Part 5/5

Humanoid Adaptation: Extending VIMA to High-DoF Humanoid Robots

19 min read

Plotjuggler G1 Debug

· 5 parts

Part 1/5

PlotJuggler + G1: ROS2 Setup & Live Topic Streaming

13 min read

Part 2/5

Visualizing 23 G1 Joints: MCAP Bags & Panel Layouts

12 min read

Part 3/5

IMU Debug: Quaternion to Euler + FFT Vibration Analysis

16 min read

Part 4/5

PlotJuggler Lua Custom Functions: Computing Power and Tracking Error for G1

9 min read

Part 5/5

ZMQ + VideoViewer: Sim-to-Real Debug for G1 with PlotJuggler

10 min read

Robot 3d Manipulation Vla

· 7 parts

Part 1/7

Why 2D RGB VLA Is Not Enough for Manipulation

12 min read

Part 2/7

Robo3R: Feed-Forward 3D Reconstruction for Manipulation

11 min read

Part 3/7

DP3: Point-Cloud 3D Diffusion Policy Hands-On

15 min read

Part 4/7

3D Perception for Whole-Body Humanoids: Beyond-FOV

7 min read

Part 5/7

WholeBodyVLA: Action-Free Video to Loco-Manipulation

6 min read

Part 6/7

Data Collection: Teleop vs Robot-Free vs Egocentric Video

6 min read

Part 7/7

A 3D Manipulation Roadmap for the Unitree G1 Humanoid

6 min read

Robot Arm Classical Controller

· 15 parts

Part 1/15

Classical Robot Arm Control: The Complete Roadmap

16 min read

Part 2/15

Setting Up a C++ Robot Controller Project with CMake

14 min read

Part 3/15

Robot Arm Model: Joints, Links, URDF, DH and SE(3)

16 min read

Part 4/15

Forward Kinematics from Scratch: DH to SE(3) in C++

14 min read

Part 5/15

Inverse Kinematics for 6-DOF Arms: Analytical vs Numerical

14 min read

Part 6/15

Jacobian-Based Control and Singularity Avoidance

14 min read

Part 7/15

Robot Coordinate Frames: Base, Tool, User and TCP

11 min read

Part 8/15

How to Implement MoveJ for Robot Arms

16 min read

Part 9/15

How to Implement MoveL: Cartesian Linear Motion with IK

13 min read

Part 10/15

MoveC and MoveP: Circular and Process Path Motion

16 min read

Part 11/15

Trapezoidal vs S-Curve Trajectory Profiles

16 min read

Part 12/15

Waypoint Blending and Corner Smoothing for Robot Arms

18 min read

Part 13/15

Robot Arm Jogging and Servoing with MoveIt Servo

15 min read

Part 14/15

Motion Planning with Collision Avoidance in MoveIt2

11 min read

Part 15/15

Executing Trajectories with ROS2 Control and UR RTDE

13 min read

Groot Sonic Humanoid Wbc

· 6 parts

Part 1/6

SONIC Architecture for Humanoid WBC

13 min read

Part 2/6

Run Sim and Evaluate SONIC Policies

13 min read

Part 3/6

BONES-SEED Data and SONIC Training

13 min read

Part 4/6

C++ Deployment: TensorRT, ZMQ, ONNX

16 min read

Part 5/6

PICO Teleop and LeRobot Data for VLA

15 min read

Part 6/6

MotionBricks Real-Time Motion Tokens

14 min read

G1 Mujoco Telemetry Control

· 6 parts

Part 1/6

Bring Up G1 MuJoCo with Low-Level DDS

13 min read

Part 2/6

Convert LAFAN1 G1 Motions to Foxglove MCAP

13 min read

Part 3/6

Debug G1 MPC/WBID with PlotJuggler

16 min read

Part 4/6

From Posture PD to Contact-Aware WBID

15 min read

Part 5/6

G1 Upper Body: IK, VR and MuJoCo

14 min read

Part 6/6

A Sim2Real Checklist for G1 Controllers

15 min read

Humanoid Data War 2026

· 7 parts

Research

Part 1/7

The Data War: Who Owns Humanoid Robot Data in 2026?

10 min read

Humanoid

Part 2/7

Teleoperation: Real-World Robot Data Collection

15 min read

Humanoid

Part 3/7

Human Video Mining: Pre-training Robots from Human Video

16 min read

Humanoid

Part 4/7

Synthetic Data Pipelines: From Sim to Real

16 min read

Part 5/7

VLA Data Scaling Laws for Robot Learning

18 min read

Humanoid

Part 6/7

Data Strategy: What Should Small Teams Collect?

18 min read

Humanoid

Part 7/7

Open vs Closed: Licenses, Data Moats, and What's Next

16 min read

VLA & LeRobot Mastery

· 17 parts

Part 1/17

LeRobot Framework Deep Dive: Architecture & API Guide

9 min read

Part 2/17

Data Collection via Teleoperation in Simulation

10 min read

Part 3/17

Training Single-Arm Policies: ACT & Diffusion Policy

9 min read

Part 4/17

Multi-Object Manipulation: Sorting, Stacking & Rearranging

8 min read

Whole-body VLA

Part 5/17

Long-Horizon Tasks: Chaining Actions for Multi-Step Manipulation

8 min read

Part 6/17

Dual-Arm Robot: Setup & Calibration with LeRobot

8 min read

Part 7/17

Bimanual Tasks: Folding, Pouring & Assembly with Dual Arms

9 min read

Part 8/17

Mobile Manipulation: Mobile Base + Arms Integration

7 min read

Whole-body VLA

Part 9/17

Humanoid Manipulation: Whole-body Control with LeRobot

7 min read

Part 10/17

Sim-to-Real Transfer: Deploying VLA Policies to Real Robots

9 min read

Part 11/17

LeRobot v0.5: What's New

10 min read

Part 12/17

SmolVLA: Train a 450M VLA on Consumer GPU

10 min read

Part 13/17

Pi0-FAST: 5x Faster Autoregressive VLA

12 min read

Part 14/17

HIL-SERL: Real Robot RL with LeRobot

16 min read

Part 15/17

PEFT/LoRA Fine-tuning & VLA Deployment

11 min read

Part 16/17

SARM in LeRobot: Reward Models for VLA

15 min read

Part 16/17

LeRobot v0.5: Pi0-FAST + G1 Whole-Body Control

11 min read

Humanoid Wbvla 2026 Lab

· 6 parts

Part 1/6

OpenWBT: G1 Teleop in MuJoCo/Isaac

13 min read

Part 2/6

TWIST2: PICO Teleop and G1 Sim2Real

16 min read

Part 3/6

EgoHumanoid: Human Demos to G1 VLA

16 min read

Part 4/6

VIRAL: RGB Sim2Real for G1 Loco-Manip

16 min read

Part 5/6

FRoM-W1: Text to Motion to G1 Policy

16 min read

Part 6/6

CLONE: MoE Teleop and Stack Choice

18 min read

Humanoid Vla Data Center 2026

· 6 parts

Part 1/6

Two-Person Pilot for Humanoid VLA Data

15 min read

Part 2/6

Choosing the Humanoid Teleop Stack

15 min read

Part 3/6

ROS 2 MCAP as the Raw Log Standard

15 min read

Part 4/6

LeRobotDataset and Robo-DM Data Lake

11 min read

Part 5/6

Synthetic Data and QA with Isaac Lab

13 min read

Part 6/6

Scaling to 20 Operators and VLA Evaluation

15 min read

Humanoid Data Ownership 2026

· 6 parts

Humanoid

Part 1/6

Humanoid Data Map 2026

15 min read

Part 2/6

VR Teleop: PICO/ZED to HDF5

18 min read

Part 3/6

Aligning Human View to Robot View

15 min read

Part 4/6

Synthetic Data with Isaac Mimic

14 min read

Part 5/6

Human Video: Phantom and pi0.5

14 min read

Part 6/6

VLA Stack: Data to Deployment

13 min read

Wholebodyvla Data Wbc

· 7 parts

Part 1/7

Mapping the WholeBodyVLA Pipeline

15 min read

Part 2/7

Ego Videos for Locomotion LAMs

17 min read

Part 3/7

Retarget AMASS and GMR to Robots

18 min read

Part 4/7

Whole-Body Teleop with TWIST/HOMIE

17 min read

Part 5/7

Training Trackers and LMO RL

16 min read

Part 6/7

Sim-to-Real Before Unitree G1

15 min read

Part 7/7

Evaluating Whole-Body VLAs

14 min read

Unifolm Vla G1 Series

· 5 parts

Part 1/5

unifolm-vla + Unitree G1 (Post 1): WBC+VLA system architecture — from data to real hardware

6 min read

Part 2/5

unifolm-vla + Unitree G1 (Post 2): collecting data with xr_teleoperate + Meta Quest 3

7 min read

Part 3/5

unifolm-vla + Unitree G1 (Post 3): data pipeline — JSON → LeRobot → HDF5 → RLDS

5 min read

Part 4/5

unifolm-vla + Unitree G1 (Post 4): fine-tuning from Qwen2.5-VL-7B — 8-GPU and single-GPU LoRA

7 min read

Part 5/5

unifolm-vla + Unitree G1 (Post 5): deploy inference server, SSH tunnel, and parallel locomotion

7 min read

Grail Humanoid Data Pipeline

· 6 parts

Part 1/6

3D Assets and Terrain in GRAIL

14 min read

Part 2/6

2D HOI Videos with Blender and Kling

15 min read

Part 3/6

4D HOI Reconstruction: GEM, SAM2, MoGe

16 min read

Part 4/6

Static-Object Mode for Terrain and Sitting

15 min read

Part 5/6

Retarget SMPL-X Motions to Unitree G1

14 min read

Part 6/6

Train SONIC, Export, and Evaluate GRAIL

14 min read

AI for Robotics

· 8 parts

Robot AI

Part 1/8

RL for Robotics: PPO, SAC and How to Choose Your Algorithm

9 min read

Robot AI

Part 2/8

Imitation Learning: BC, DAgger and DAPG for Robots

11 min read

Part 3/8

Action Chunking Transformers (ACT): Architecture in Detail

11 min read

Part 4/8

Diffusion Policy: A Revolution in Robot Manipulation

9 min read

Robot AI

Part 5/8

VLA Models: RT-2 → Octo → OpenVLA → π0

12 min read

Part 6/8

SpatialVLA: 3D Understanding for Robot Manipulation

8 min read

Part 7/8

Hands-on: Fine-tune OpenVLA with LeRobot

13 min read

Part 8/8

OpenVLA: Open VLA for Robots

15 min read

GR00T Whole-body VLA Data Pipeline

· 4 parts

Part 1/4

GR00T Whole-Body VLA Data: Open Datasets

13 min read

Part 2/4

GR00T Whole-Body VLA Data: Simulation Data

13 min read

Part 3/4

GR00T Whole-Body VLA Data: Do You Need Real Data?

13 min read

Part 4/4

GR00T Whole-Body VLA: Training SONIC Controller

8 min read

Vla Wbc Repos

· 3 parts

Humanoid

Part 1/3

VLA + WBC repos landscape 2025-2026: all humanoid GitHub repos mapped

5 min read

Humanoid

Part 2/3

VLA + WBC repos from the US: NVIDIA GR00T, openpi, HumanPlus, TeleVision

8 min read

Part 3/3

VLA + WBC repos from China: Unitree, THU RDT-1B, and the open community

8 min read

Umi Vla Series

· 7 parts

Part 1/7

What is UMI? How to collect VLA robot data without teleop

6 min read

Part 2/7

Build your first UMI gripper: 3D printing, assembly, and GoPro smoke test

7 min read

Part 3/7

Collect single-arm UMI demos and run the official SLAM pipeline

7 min read

Part 4/7

Train your first UMI Diffusion Policy and test on a real robot arm

5 min read

Part 5/7

Go bimanual: UMI two-arm pipeline with official scripts

7 min read

Manipulation

Part 6/7

Upgrading to D405: when to replace GoPro in UMI and how

6 min read

Part 7/7

Whole-body VLA: combining UMI + mocap/VR for full-body humanoid data

7 min read

Groot N1 G1 Series

· 5 parts

Part 1/5

GR00T N1 + Unitree G1: decoupled WBC+VLA architecture from 6Hz to 500Hz

6 min read

Part 2/5

GR00T N1 + G1 (Post 2): data collection in Isaac Lab and xr_teleoperate → LeRobot

5 min read

Part 3/5

GR00T N1 + G1 (Post 3): fine-tuning GR00T N1 — GPU, config, training script

5 min read

Part 4/5

GR00T N1 + G1 (Post 4): deploying GR00T-WBC on Unitree G1 — GEAR + SONIC

6 min read

Part 5/5

GR00T N1 + G1 (Post 5): sim2real transfer, domain randomization, and eval with humanoid-bench

6 min read

Simplevla Rl

· 11 parts

Part 1/11

SimpleVLA-RL (1): Overview & Key Ideas

9 min read

Part 2/11

SimpleVLA-RL (2): Architecture & Algorithm

14 min read

Part 3/11

SimpleVLA-RL (3): Setup & Training

11 min read

Part 4/11

SimpleVLA-RL (4): Results & Key Takeaways

12 min read

Whole-body VLA

Part 5/11

SimpleVLA-RL (5): Comparison with LeRobot

11 min read

Part 6/11

SimpleVLA-RL (6): OpenArm — Training Roadmap

12 min read

Whole-body VLA

Part 7/11

SimpleVLA-RL (7): Collecting Data for OpenArm

14 min read

Part 8/11

SimpleVLA-RL (8): Training & Deploying on OpenArm

13 min read

Part 9/11

SimpleVLA-RL (9): OpenArm Simulation & Data Collection

17 min read

Part 10/11

SimpleVLA-RL (10): SFT & RL Training for OpenArm

15 min read

Part 11/11

SimpleVLA-RL (11): Sim-to-Real Transfer for OpenArm

16 min read

Psi0 Hands On

· 6 parts

Part 1/6

Ψ₀ Hands-On (1): Overview & Key Ideas

12 min read

Part 2/6

Ψ₀ Hands-On (2): The Three-Tier Architecture

15 min read

Part 3/6

Psi0 Hands-On (3): Data Recipe & Pipeline

12 min read

Part 4/6

Psi0 Hands-On (4): Setup & Training Pipeline

14 min read

Part 5/6

Psi0 Hands-On (5): Inference & Evaluation

13 min read

Part 6/6

Psi0 Hands-On (6): Ablation Studies & Key Takeaways

15 min read

RL Locomotion for Humanoid Robots

· 10 parts

Part 1/10

RL Locomotion for Humanoids: Fundamentals & Environment Setup

10 min read

Part 2/10

Reward Engineering for Bipedal Walking: The Art of Reward Design

10 min read

Part 3/10

Unitree G1: Training a Walking Policy from Scratch in Isaac Lab

9 min read

Part 4/10

Unitree G1: Terrain Adaptation & Robust Walking

8 min read

Part 5/10

Unitree H1: Full-size Humanoid Locomotion Training

8 min read

Part 6/10

Unitree H1: Running, Turning & Dynamic Motions

8 min read

Part 7/10

Unitree H1-2: Enhanced Locomotion with New Hardware

7 min read

Part 8/10

Loco-Manipulation: Walking While Carrying & Manipulating Objects

7 min read

Part 9/10

Humanoid Parkour: Jumping, Climbing & Obstacle Courses

7 min read

Humanoid

Part 10/10

Sim-to-Real for Humanoids: Deployment Best Practices

9 min read

RL for Robot Manipulation

· 8 parts

Manipulation

Part 1/8

RL for Manipulation: MDP, Reward Design & Environment Setup

12 min read

Manipulation

Part 2/8

Grasping with RL: Stable Grasp & Object Variety

11 min read

Part 3/8

RL Force Control: Balancing a Cup of Water

10 min read

Manipulation

Part 4/8

Precision Pick-and-Place: Position & Orientation Control

7 min read

Manipulation

Part 5/8

Carrying & Transporting Objects: Stability During Motion

8 min read

Part 6/8

Contact-Rich Manipulation: Assembly, Insertion & Peg-in-Hole

7 min read

Manipulation

Part 7/8

Tool Use: Teaching Robots to Use Tools with RL

7 min read

Manipulation

Part 8/8

Multi-Step Manipulation: Curriculum Learning for Long-Horizon Tasks

8 min read

Simulation for Robotics

· 6 parts

Simulation

Part 1/6

Simulation for Robotics: MuJoCo vs Isaac Sim vs Gazebo

9 min read

Part 2/6

Getting Started with MuJoCo: From Installation to First Robot Simulation

10 min read

Part 3/6

NVIDIA Isaac Lab: GPU-Accelerated RL Training from Zero

11 min read

Part 4/6

Domain Randomization: The Key to Sim-to-Real Transfer

9 min read

Simulation

Part 5/6

Sim-to-Real Pipeline: From Training to Real Robot

15 min read

Simulation

Part 6/6

Digital Twins and ROS 2: Simulation in Manufacturing

10 min read

ROS 2 A to Z

· 7 parts

Part 1/7

ROS 2 A to Z Part 1: Setup and First Node

9 min read

Part 2/7

ROS 2 A to Z (Part 2): Topics, Services, and Actions

9 min read

Part 3/7

ROS 2 A to Z (P3): Parameters, Launch, Lifecycle

7 min read

Part 4/7

ROS 2 A to Z (P4): TF2, URDF and RViz2

7 min read

Part 5/7

ROS 2 A to Z (P5): ros2_control and Hardware

11 min read

Part 6/7

ROS 2 A to Z (P6): Nav2 — Autonomous Navigation

10 min read

Part 7/7

ROS 2 A to Z (P7): Simulation, rosbag2 and Debug

8 min read

Robot Manipulation Masterclass

· 7 parts

Part 1/7

Robot Grasping 101: Analytical to Learning-Based

8 min read

Part 2/7

Imitation Learning for Manipulation: BC, DAgger, ACT

8 min read

Part 3/7

Diffusion Policy in Practice: From Theory to Code

7 min read

Part 4/7

VLA for Manipulation: RT-2, Octo, pi0 in Practice

8 min read

Part 5/7

Dexterous Manipulation: Teaching Robot Hands

7 min read

Part 6/7

Bimanual Manipulation: Teaching Robots to Use Both Arms

7 min read

Part 7/7

Building Manipulation Systems with LeRobot: From Zero to Deployment

8 min read

Locomotion: Zero to Hero

· 7 parts

Part 1/7

Locomotion Basics: From ZMP to CPG

11 min read

Part 2/7

RL for Locomotion: PPO, Reward Shaping and Curriculum

10 min read

Part 3/7

Quadruped Locomotion: legged_gym to Unitree Go2

10 min read

Part 4/7

Walk These Ways: Adaptive Locomotion One Policy

9 min read

Part 5/7

Robot Parkour: Jumping and Climbing with RL

9 min read

Part 6/7

Bipedal Walking: Controlling 2-Legged Robots with RL

10 min read

Part 7/7

Sim-to-Real for Locomotion: Reality and Experience

12 min read

Humanoid Robot Engineering

· 6 parts

Humanoid

Part 1/6

Humanoid Robots 2026: Complete Platform Overview

8 min read

Humanoid

Part 2/6

Humanoid Control Basics: IK to Task-Space Control

8 min read

Part 3/6

Whole-Body MPC: Real-Time Full-Body Control

6 min read

Part 4/6

RL for Humanoid: From Humanoid-Gym to Sim2Real

8 min read

Part 5/6

Loco-Manipulation: Walking and Manipulating Simultaneously

8 min read

Humanoid

Part 6/6

The Future of Humanoids: Opportunities for Robotics Engineers

9 min read

Modern Navigation

· 5 parts

Part 1/5

SLAM A to Z: LiDAR, Visual and How Robots Localize

8 min read

Part 2/5

ROS 2 Nav2: Complete Navigation for AMR

7 min read

Part 3/5

Learning-based Navigation: GNM, ViNT and NoMaD

8 min read

Part 4/5

Vision-Language Navigation: Robot Following Instructions

8 min read

Part 5/5

Outdoor Navigation and Multi-Robot Coordination

10 min read

NEWTutorial

StarVLA-WBCUnitree G1VLA

wholebody-vla

StarVLA-WBC: train G1 whole-body VLA

A practical StarVLA-WBC guide: train a Unitree G1 whole-body VLA on SIMPLE and deploy it through a WBC/SONIC bridge.

9/11/202614 min read

Tutorial

VLA-Precision: ACoB Online RL in 45 Minutes

vlaonline-rlacob

wholebody-vla

VLA-Precision: ACoB Online RL in 45 Minutes

A practical guide to VLA-Precision, ACoB online RL, ACoB-Stream, setup, training, inference, and lab manipulation results.

9/9/202615 min read

Tutorial

ns-vlavlalibero

wholebody-vla

NS-VLA v2: Fine-tune VLA on LIBERO

A practical NS-VLA v2 guide: neuro-symbolic primitives, BC warmup, GRPO/AWR fine-tuning, inference, and LIBERO results.

9/7/202612 min read

Tutorial

FlashVLAπ0.5SmolVLA

wholebody-vla

FlashVLA: Speed Up π0.5/SmolVLA

A practical FlashVLA guide for π0.5 and SmolVLA: streaming action decoding, setup, training, async inference, and benchmark results.

9/4/202613 min read

Tutorial

guided-action-flowsmolvlaq-guided-flow

wholebody-vla

Guided Action Flow for SmolVLA

A beginner guide to GAF: improve frozen SmolVLA on LIBERO with a Q-guided critic, without policy fine-tuning.

9/2/202615 min read

Tutorial

ma-vlavlalerobot

wholebody-vla

MA-VLA: Multi-Arm VLA on LeRobot

A beginner guide to MA-VLA with atomic actions, Arm Shuffle, LeRobot data, training, inference, and benchmark results.

8/31/202613 min read

Tutorial

gvlamigavla

wholebody-vla

Train GVLA with MiGA for Any Gripper

A beginner-friendly guide to GVLA, MiGA 103K demos, gripper-aware fine-tuning, inference, and results across multiple robot grippers.

8/28/202613 min read

Tutorial

maniguardvlaopenpi

wholebody-vla

ManiGuard-Bench for Safe VLA Eval

A beginner guide to evaluating safe VLA manipulation with ManiGuard-Bench, openpi, GR00T, and SmolVLA.

8/26/202612 min read

Tutorial

PoseVLApi0.5RoboTwin

wholebody-vla

PoseVLA: 3D pose pretraining for π0.5

A practical PoseVLA guide covering 3D pose pretraining, π0.5 branches, RoboTwin post-training, inference, and the 89% release result.

8/24/202614 min read

Tutorial

EvoVLA ECCV 2026: Killing Stage Hallucination in VLA

vlamanipulationreinforcement-learning

manipulation

EvoVLA ECCV 2026: Killing Stage Hallucination in VLA

EvoVLA (ECCV 2026) guide: eliminating stage hallucination in VLA manipulation via Stage-Aligned Reward, Gemini hard negatives, and long-horizon memory to reach

8/21/202610 min read

Tutorial

VLASH: Real-Time VLAs via Async Inference (11.8× Faster)

vlareal-timeasynchronous-inference

wholebody-vla

VLASH: Real-Time VLAs via Async Inference (11.8× Faster)

MIT Han Lab's VLASH makes VLAs real-time via future-state-aware asynchronous inference — 11.8× lower reaction latency, no architecture changes, open-source.

8/20/202612 min read

Tutorial

GEAR-SONIC Science Robotics 2026: Scaling 100M Frames for Humanoid

humanoidwhole-body-controlnvidia

wholebody-vla

GEAR-SONIC Science Robotics 2026: Scaling 100M Frames for Humanoid

Complete guide to training SONIC whole-body controller with GR00T-WholeBodyControl open-source on 100M+ motion capture frames, deployed on Unitree G1.

8/19/202612 min read

Blog

Latest Posts

FluxVLA Engine: Train, Eval, Deploy VLA

TacPAC: Tactile Action-Chunk Correction

Show-Harness: VLMs Play Robots with GUMI

Browse Series

cuRobo on Jetson AGX Orin and Unitree G1

cuRobo on Jetson AGX Orin 64GB and Unitree G1: a 10-part roadmap

Installing cuRoboV2 on Jetson AGX Orin 64GB: environment, CUDA, and smoke tests

Building a cuRobo robot model: URDF, collision spheres, and self-collision

FK and IK in cuRobo: checking frames, quaternions, and joint order

Motion planning and grasping with cuRobo: obstacles, seeds, and trajectories

A ROS 2 bridge for cuRobo on Jetson: from goal pose to JointTrajectory

Scene mapping for cuRobo: depth cameras, TSDF/ESDF, and obstacle snapshots

Unitree G1 arm-only planning with cuRobo: moving arms without breaking balance

Unitree G1 whole-body with cuRobo: floating base, tool frames, and retargeting

Deploying cuRobo on real robots: checklist, current issues, and rollback

Ai Manipulation Agents

Why Multi-Agent Pipelines Beat Single VLA Models | AI Manipulation Agents #1

Perception Agent: Florence-v2 + AnyGrasp | AI Manipulation Agents #2

ALRM: Code-as-Policy vs Tool-as-Policy in ReAct | AI Manipulation Agents #3

Agentic Robot: SAP Protocol & Temporal Verifier

Sim-to-Real Deploy: Moving the SAP Pipeline from LIBERO to Physical Robots | AI Manipulation Agents #5

Vima Humanoid Manip

VIMA Architecture: Cross-Attention Transformer Explained

VimaBench: 17 Tasks and 4-Level Generalization Protocol

Object Tokenizer: From Raw Pixels to Object Tokens with Mask R-CNN + ViT

Dataset 650K: Collecting Large-Scale Multi-Task Manipulation Data for VIMA

Humanoid Adaptation: Extending VIMA to High-DoF Humanoid Robots

Plotjuggler G1 Debug

PlotJuggler + G1: ROS2 Setup & Live Topic Streaming

Visualizing 23 G1 Joints: MCAP Bags & Panel Layouts

IMU Debug: Quaternion to Euler + FFT Vibration Analysis

PlotJuggler Lua Custom Functions: Computing Power and Tracking Error for G1

ZMQ + VideoViewer: Sim-to-Real Debug for G1 with PlotJuggler

Robot 3d Manipulation Vla

Why 2D RGB VLA Is Not Enough for Manipulation

Robo3R: Feed-Forward 3D Reconstruction for Manipulation

DP3: Point-Cloud 3D Diffusion Policy Hands-On

3D Perception for Whole-Body Humanoids: Beyond-FOV

WholeBodyVLA: Action-Free Video to Loco-Manipulation

Data Collection: Teleop vs Robot-Free vs Egocentric Video

A 3D Manipulation Roadmap for the Unitree G1 Humanoid

Robot Arm Classical Controller

Classical Robot Arm Control: The Complete Roadmap

Setting Up a C++ Robot Controller Project with CMake

Robot Arm Model: Joints, Links, URDF, DH and SE(3)

Forward Kinematics from Scratch: DH to SE(3) in C++

Inverse Kinematics for 6-DOF Arms: Analytical vs Numerical

Jacobian-Based Control and Singularity Avoidance

Robot Coordinate Frames: Base, Tool, User and TCP

How to Implement MoveJ for Robot Arms

How to Implement MoveL: Cartesian Linear Motion with IK

MoveC and MoveP: Circular and Process Path Motion

Trapezoidal vs S-Curve Trajectory Profiles

Waypoint Blending and Corner Smoothing for Robot Arms

Robot Arm Jogging and Servoing with MoveIt Servo

Motion Planning with Collision Avoidance in MoveIt2

Executing Trajectories with ROS2 Control and UR RTDE

Groot Sonic Humanoid Wbc

SONIC Architecture for Humanoid WBC

Run Sim and Evaluate SONIC Policies

BONES-SEED Data and SONIC Training

C++ Deployment: TensorRT, ZMQ, ONNX

PICO Teleop and LeRobot Data for VLA

MotionBricks Real-Time Motion Tokens

G1 Mujoco Telemetry Control

Bring Up G1 MuJoCo with Low-Level DDS

Convert LAFAN1 G1 Motions to Foxglove MCAP

Debug G1 MPC/WBID with PlotJuggler

From Posture PD to Contact-Aware WBID

G1 Upper Body: IK, VR and MuJoCo

A Sim2Real Checklist for G1 Controllers

Humanoid Data War 2026

The Data War: Who Owns Humanoid Robot Data in 2026?

Teleoperation: Real-World Robot Data Collection

Human Video Mining: Pre-training Robots from Human Video

Synthetic Data Pipelines: From Sim to Real

VLA Data Scaling Laws for Robot Learning

Data Strategy: What Should Small Teams Collect?