Bridging the Gap Between Neural Networks and Actuation
For years, humanoid robotics developers have faced a frustrating physical reality. Software intelligence advances at breakneck speeds, scaling parameters in large language models and neural control policies by orders of magnitude. Yet, the physical hardware executing those commands often stumbles. Centralized controllers introduce latency, round-trip processing lags, and compliance failures that keep robots confined to controlled laboratory environments.
HIGEN RNM targets this exact engineering friction point. By decentralizing intelligence down to the actuator level, the company’s new architecture allows each individual robotic joint to detect unexpected physical disruptions independently. The system adjusts local force and stiffness in less than 100 milliseconds. It bypasses the traditional bottleneck of routing every minor sensory correction back through a central computing unit.
Vertical Integration and High-Density Engineering
Moving from a custom research prototype to volume manufacturing requires strict component standardization. To achieve this, HIGEN RNM has engineered six distinct modular actuator platforms ranging from 60 Nm to 348 Nm. These units scale across an entire humanoid form factor, addressing high-torque lower-body requirements like hips and knees while fitting into tighter spatial constraints like wrists and necks.
The architectural secret sauce lies in deep vertical integration. The system bundles an axial flux permanent magnet (AFPM) motor, a proprietary 3K compound planetary gearbox, dual encoders, and integrated drive electronics into a single, cohesive package.
Mechanical compliance is equally critical when machines operate alongside human workers. The proprietary 3K compound planetary gearbox achieves a back-drive torque of under 1 Nm under defined testing conditions. This low mechanical impedance ensures that if a robotic arm makes unexpected contact with a human operator, the joint yields naturally rather than pushing through the collision with rigid, unyielding force.
Eliminating Dedicated Torque Sensors
Cost and hardware complexity remain major hurdles in scaling humanoid hardware. Traditional high-performance joints often rely on dedicated, expensive joint-torque sensors that introduce additional points of failure and wiring complexity. HIGEN RNM bypasses this hardware overhead through clever firmware and electronic design.
By leveraging real-time data streaming from the dual encoders alongside precise measurements of motor current, the system estimates external forces algorithmically. Developers get high-fidelity force feedback without needing a dedicated torque sensor at every single articulation point. This approach trims bill-of-materials costs and reduces physical wiring harnesses inside slender robotic limbs.
Safety Certification and Commercial Roadmaps
Scaling industrial and service robotics demands rigorous compliance with international safety standards. HIGEN RNM has baked ISO 13849-1 functional safety requirements directly into its development pipeline. The company is actively pursuing CE and UL certifications for its actuator modules, paving the way for international commercial deployments.
Three distinct Korean patent applications covering advanced sensor technology and adaptive force control algorithms back the platform’s core intellectual property. HIGEN RNM plans to showcase the Human-Friendly Actuator Platform, alongside live performance benchmarks and application demonstrations, on the global stage at CES 2027 in Las Vegas.
The 30-Second Verdict
- Core Innovation: Decentralized joint control reacting to disruptions in under 100 ms.
- Power & Size: AFPM-based motor architecture increases torque density by 30% while reducing volume by 30%.
- Safety: Sub-1 Nm back-drive torque supports safe human-robot collaboration, backed by ISO 13849-1 compliance efforts.
- Public Debut: Set for live demonstrations at CES 2027.
