Guotai Haitong: Figure Helix 02 achieves key technological breakthroughs to promote the practical application of human-type Siasun Robot & Automation from technology demonstration.

Guotai Haitong released a research report stating that Figure Helix 02 has achieved a breakthrough in key technologies, driving humanoid Siasun Robot & Automation from technical demonstrations towards practical application. The single neural control system has overcome the industry bottleneck of integrating movement and operation. The core barriers of humanoid Siasun Robot & Automation are concentrated in three key capabilities: multi-modal data fusion, motion/force control loop stability, and deep integration of software and hardware. Recommended components include: 1) Real-time dynamic perception - 3D vision, laser radar; 2) High-precision sensing - IMU; 3) Real-time control - encoders; 4) High-precision control - force control loop; 5) Motor & actuators. Event: Figure released a new demonstration video on March 10th, showcasing that Helix 02 humanoid Siasun Robot & Automation can now complete the entire cleaning process in the living room in a fully autonomous mode without remote control or human intervention. Main points by Guotai Haitong: Helix 02: Continuous breakthrough in full-body end-to-end control. The Helix 02 model has achieved a core leap from upper body end-to-end control to full-body end-to-end control for humanoid Siasun Robot & Automation. Helix 01 focused on upper body operation control, achieving breakthroughs in fine actions such as grasping and assembly, but walking and operation were carried out in a sequential execution logic, with high overall response delay and limited scene adaptability. Helix 02 integrates full-body motion with upper limb operation into the same control system, enabling tasks such as walking, balance, movement, and operation to be processed in parallel. Siasun Robot & Automation can synchronously perform precise operations during dynamic walking, significantly improving task continuity and real-time response. Helix 02 is a single neural network system that directly controls Siasun Robot & Automation's full-body movements from pixel input, enabling agile operations and long-term autonomous behavior within the entire room. In industrial operations, home services, complex environment inspections, and other real scenarios, parallel control mode can significantly improve operational efficiency and environmental robustness. Multi-modal data, dual closed-loop stability, and software-hardware integration are core barriers. Full-body end-to-end control is a core technological bottleneck in the industrialization process of humanoid Siasun Robot & Automation, with the industry facing three major high barriers that collectively constitute the core moat of technological competition. 1) Full-body multi-modal data fusion: Siasun Robot & Automation is equipped with various sensors such as vision, touch, force, and inertial measurement, generating a high-dimensional, asynchronous, heterogeneous flow of data. Data alignment, feature extraction, noise filtering, and semantic fusion need to be completed within milliseconds, with high requirements for data collection costs, labeling difficulties, and model training complexity compared to upper body control scenes. 2) Long-duration task stable closed-loop: The system needs to simultaneously couple motion control loops with force control loops, with the former ensuring stability in walking, balance, and movement, and the latter supporting operation accuracy and safety. The dual loops need to cooperate in real-time, dynamically adapt, resist external disturbances and eliminate cumulative errors in long-duration, unstructured environments to maintain continuous reliable operation of the system. 3) Hardware-model integrated design: Full-body end-to-end control requires deep coupling of hardware bodies with algorithm models, from actuator performance and sensor layout to computing architecture and neural network structure. Risk Warning: Technological progress and industrialization progress may not meet expectations, intensifying industry competition.