Helix 02 Dynamic Whole-Body Control

6/30/2026

This post introduces the application of Helix 02, Figure's VLA, to power Figure 03's dynamic whole-body control in a manufacturing sequencing use case. It highlights Helix 02's ability to coordinate hands, arms, torso, and feet for loco-manipulation, enabling the robot to grasp parts while balancing, repositioning, and precisely placing them. The post details how Helix 02 continuously adapts movements to spatial variations and corrects errors for tasks that require both manipulation and locomotion, such as pulling a heavy cart.

5/8/2026

This post demonstrates a significant advancement in the Helix 02 system by showcasing multi-humanoid collaborative locomanipulation. The key technical contribution is the ability for two Helix-02 robots to perform complex tasks like making a bed and tidying a bedroom together, directly from pixels to actions using a single learned neural network. This is achieved without explicit inter-robot communication or a central planner, relying instead on each robot inferring the partner's intent from motion alone. The post highlights specific behaviors enabled by this collaborative capability, including coordinated manipulation of deformable objects (bedding) and shared workspaces, as well as individual dexterous tasks performed in a collaborative context (e.g., operating a trash can pedal while balancing).

4/29/2026

This post details the significant scaling of Figure 03 robot production, achieving a 24x throughput increase to one robot per hour. It highlights improvements in manufacturing quality, yield rates, and end-of-line testing. Operationally, it describes the development of robust diagnostics, fallback ladders, and a custom Fleet Management System with OTA updates. A key technical breakthrough is the integration of perception-conditioned whole-body control into Helix's System 0 (S0), enabling the AI to process visual input for environmental awareness and perform complex maneuvers like stair traversal via zero-shot sim-to-real transfer.

3/9/2026

Demonstrates Helix 02's ability to perform whole-body, end-to-end living room cleanup by learning new tasks from data, including coordinated tool use (spray bottle and towel wiping), dynamic handling of flexible objects (towels), complex bimanual manipulation (scooping blocks into a bin), whole-body strategies for efficiency (tucking a container under an arm), dynamic object throws (pillows), in-hand reorientation (remote control), reorganizing tools during motion, and navigating tight spaces with precise foot placement.

1/27/2026

Introduces Helix 02, a unified whole-body loco-manipulation VLA that extends control to the entire robot. Key contributions include: 1. System 0, a learned whole-body controller trained on 1000+ hours of human motion data and sim-to-real RL, replacing 109,504 lines of hand-engineered C++ with a single neural prior for stable motion. 2. System 1's expansion to connect all sensors (head cameras, palm cameras, fingertip tactile sensors, proprioception) to all actuators (full-body joint control), enabling dexterous manipulation with new hardware modalities. 3. Demonstration of long-horizon autonomous tasks (e.g., 4-minute dishwasher loading/unloading) and new classes of dexterity (e.g., pill extraction, syringe dispensing) enabled by the integrated system.

11/19/2025

Details the deployment of Figure 02 at BMW for sheet-metal loading, including KPIs (cycle time, placement accuracy, interventions) and technical challenges. Highlights hardware reliability learnings, specifically the re-architecture of the forearm subsystem to improve reliability and simplify thermal management for Figure 03.

10/9/2025

Introduced Figure 03, a third-generation humanoid robot with a ground-up hardware and software redesign. Key advancements include a redesigned sensory suite and hand system for the Helix AI, featuring a next-generation vision system with doubled frame rate, quartered latency, and a 60% wider field of view per camera, plus an embedded palm camera for redundant visual feedback. New compliant and tactile hand designs with a first-generation tactile sensor capable of detecting 3-gram forces are integrated. The robot supports 10 Gbps mmWave data offload. For home use, it incorporates safety features, an upgraded audio system, and wireless inductive charging. The design is optimized for mass manufacturing using tooled processes and a new supply chain, with the BotQ facility capable of producing 12,000 robots annually. Commercial applications benefit from faster actuators, improved torque density, and enhanced perception for precise manipulation.

9/18/2025

Introduced Project Go-Big for internet-scale humanoid pretraining using a partnership with Brookfield to collect diverse real-world human video data. Achieved zero-shot human video-to-robot transfer for navigation, enabling Helix to navigate complex environments from natural language commands using only human video data. Integrated navigation capabilities into a single Helix network that also handles manipulation, eliminating the need for separate systems.

9/3/2025

Demonstrates the adaptability of the Helix VLA model to new, complex robotic tasks like dishwasher loading by ingesting new data, without requiring new algorithms or special-case engineering. Highlights capabilities in object singulation, precise manipulation with reorientation, adaptation to varied starting states, and error recovery.

8/12/2025

The Helix VLA model architecture was applied to the new task of laundry folding without modifications. This involved training on a new dataset specifically for this task, demonstrating the model's ability to generalize to deformable object manipulation. Key capabilities demonstrated include picking from a mixed pile, adaptive folding strategies, error recovery, and fine manipulation skills like edge tracing and corner pinching. The system operates end-to-end, directly mapping vision and language input to motor control, bypassing explicit object representations for deformable items.

6/7/2025

This post details significant advancements in Helix's logistics capabilities, focusing on scaling its performance through data and architectural improvements. Key contributions include the integration of a temporal vision memory module for stateful perception and context awareness, the incorporation of force feedback for more precise manipulation and touch sensing, and the augmentation of the policy with a history of recent states for faster, more reactive control. These enhancements have led to handling a wider variety of package types (deformable bags, flat envelopes), increased throughput to 4.05 seconds per package, and improved barcode scanning success to ~95%. The post also quantifies the impact of scaling demonstration training data from 10 to 60 hours, showing steady performance gains.

2/26/2025

This post details significant improvements to Helix's System 1 (visuo-motor control policy) for real-world logistics applications. Key contributions include: implementing implicit stereo vision for richer 3D understanding and precise depth-aware motion; developing multi-scale visual representation for capturing fine-grained details and scene-level understanding; enabling learned visual proprioception for self-calibration and seamless cross-robot transfer; and introducing a 'Sport Mode' for test-time speed up of policy execution. The post also discusses the trade-offs between data quality and quantity, showing that curated data yields better results, and quantifies performance improvements through metrics like normalized effective throughput.

2/20/2025

Introduced Helix, a generalist Vision-Language-Action (VLA) model for humanoid control. Key contributions include: enabling full-upper-body control (35-DoF at 200Hz), demonstrating multi-robot collaboration on novel tasks, achieving zero-shot object manipulation via natural language prompts, and implementing a 'System 1, System 2' architecture for decoupled high-level understanding and low-level control. The model is trained end-to-end with a single set of weights and optimized for onboard embedded GPU deployment.