On 31 August HIGEN RNM unveiled the Human-Friendly Actuator Platform, a family of six standardized humanoid joint actuators from 60 to 348 Nm. Each joint carries its own sensing and control, so it corrects force in under 100 milliseconds without a round trip to the central controller. This is a platform announcement, not a deployment: performance data is promised for CES 2027.
Key takeaways
- Six standardized actuators from 60 to 348 Nm, covering wrists, necks, knees and hips
- Disturbance detection plus force and stiffness correction in under 100 ms, locally at the joint
- An AFPM motor, a 3K compound planetary gearbox, dual encoders and embedded drive electronics
- A claimed 30% cut in actuator volume alongside a 30% rise in torque density
- Backdriving torque below 1 Nm under the manufacturer’s specified test conditions
The bottleneck is the joint, not the brain
Vision-Language-Action models and Physical AI at large keep improving perception, reasoning and planning. The trouble starts lower down: friction, heat, mechanical nonlinearities and joint latency still separate simulation from the shop floor. HIGEN RNM attacks that layer by putting sensing and control directly inside every joint.
AI is making the robot’s brain dramatically smarter, but the point where that intelligence ultimately meets the physical world is the joint.
Jae Hak Kim, CEO of HIGEN RNM.
What sits inside one joint
The platform is unified: the same mechanical and electronic stack scales across six torque variants. The company vertically integrates the AFPM motor, gearbox and drive electronics.
The stack inside one HIGEN RNM actuator:
| Parameter | Value |
|---|---|
| Torque range | 60–348 Nm across six variants |
| Applications | wrists, necks, knees, hips |
| Joint response time | under 100 ms |
| Backdriving torque | below 1 Nm |
| Volume / torque density | −30% volume, +30% density |
| Functional safety | ISO 13849-1 in the design process, CE and UL pending |
Force sensing without a torque sensor
A conventional collaborative joint carries a dedicated torque sensor — expensive and bulky. HIGEN RNM estimates external force from dual-encoder data and motor current instead, with no separate sensor. The company also reports that its 3K gearbox reaches backdriving?backdriving: The resistance a gearbox puts up when a human moves the robot’s arm. The lower it is, the more easily the joint gives way. torque below 1 Nm, which is meant to give mechanical compliance around people.
Certification pending, product not yet shipping
The company has filed three Korean patent applications covering its sensing and adaptive force-control technologies. ISO 13849-1 functional safety requirements were built into the design process, while CE and UL certification is still under way. HIGEN RNM plans to show the platform, performance data and application demos at CES 2027 in Las Vegas.
Why it matters
A humanoid has dozens of joints, so the cost of a single one decides whether volume production adds up at all. Dropping the separate torque sensor in favour of current-and-encoder estimation is a scalability decision, not a spec-sheet record. The usual price is coarser resolution at low forces — and that is exactly what is worth checking at CES.
What's next?
- The platform, performance data and application demos are set for CES 2027 in Las Vegas
- CE and UL certification is still in progress, so conformity is not yet confirmed
- Three Korean patent applications are awaiting examination
Sources
- Robotics Business News — HIGEN RNM Unveils Human-Friendly Actuator Platform for Safer, Scalable Humanoid Robots
- RoboticsTomorrow — HIGEN RNM Targets Humanoid Robotics' Next Bottleneck: The Actuator
