As Vision-Language-Action models grow more sophisticated, the bottleneck for humanoid robots has shifted from software intelligence to mechanical latency. Real-world variables like heat, friction, and nonlinearities often derail the performance seen in simulations. By embedding sensing and control directly into the hardware, HIGEN RNM aims to bypass the delays inherent in round-trip processing through a central controller.
The new architecture features six standardized joints, ranging from 60 Nm to 348 Nm, suitable for everything from delicate wrists to load-bearing hips. Each unit integrates AFPM motors, dual encoders, and 3K compound planetary gearboxes. By utilizing dual-encoder data and motor-current sensing, the system estimates external forces without needing dedicated torque sensors. This integration allows for a 30% reduction in volume alongside a 30% increase in torque density.
CEO Jae Hak Kim emphasizes that the joint is where intelligence finally meets the physical world. To ensure safety in human-centric environments, the company has incorporated ISO 13849-1 functional safety requirements and achieved backdriving torque below 1 Nm. With three patents pending and international certifications underway, HIGEN RNM plans to demonstrate the platform's capabilities at CES 2027 in Las Vegas.

Comments (0)
No comments yet. Be the first!