The RHU series is built specifically for the bionic joints of humanoid robots, focusing on light-load, high-frequency commercial scenarios. With an ultra-compact structure and high power density at its core, it delivers strong power output and flexible motion control in limited space, helping robots achieve a lighter body structure, more natural movement, and more stable operation — providing efficient, reliable power for the new generation of humanoid robots.
| Operating Voltage | 24–48V | Encoder Type | Dual Absolute |
|---|---|---|---|
| Operating Noise | <60dB | Encoder Accuracy | 19-bit |
| Operating Temperature | -20~60℃ | Power-off Brake | Optional |
| Service Life | >10,000h | Communication Protocol | EtherCAT/CANopen |
It offers both terminal-block and flying-lead standard interfaces, supporting flexible electrical integration of the joint module and meeting the dual demands for wiring freedom and connection reliability inside the compact joints of humanoid robots.
Very high torque output from a small body, with a clear lead in power density. It helps robot joints achieve lighter weight and smaller size, providing stronger power in limited space and improving overall motion performance and endurance.
It integrates a standards-compliant Safety Torque Off (STO) function and an electromagnetic brake, together forming a dual safety system for humanoid robot joints — comprehensive protection from the signal layer to the physical layer.
The RHU32H and RHU40H models offer a 22mm hollow bore — the largest in the RHU lineup. This matters because humanoid robots route a large number of cables through each joint: power lines, EtherCAT communication cables, encoder wires, torque sensor signals, and sometimes pneumatic tubing. A 22mm bore can accommodate all of these simultaneously, eliminating external cable routing that would otherwise restrict joint range of motion, add weight, and create interference risks during dynamic movement.
The RHU-F is a force-control variant that adds a high-precision torque sensor integrated at the harmonic reducer output. Key specs: torque accuracy ≤0.5% FS, sensor resolution 0.1% FS, sampling frequency ≥5kHz. It supports EtherCAT and CAN FD protocols. Four models are available: RHU14L-F, RHU17L-F, RHU20L-F, and RHU25L-F. The RHU-F enables force-controlled interaction, compliant motion, impedance control, and collision detection — capabilities critical for humanoid robots operating in human environments.
The standard RHU series supports EtherCAT and CANopen. The RHU-F force-control variant supports EtherCAT and CAN FD. EtherCAT provides real-time, high-speed, synchronized multi-axis control required by professional humanoid robot platforms. CANopen is suitable for multi-joint daisy-chain setups. Protocol switching between CANopen and CAN custom is available by contacting the manufacturer.
The RHU series includes 6 standard models. Outer diameters range from 52mm (RHU14L) to 130mm (RHU40H). Reduction ratios span 50:1 to 160:1. Rated torque at 2000rpm ranges from 3.7 N.m (RHU14L-52) to 363 N.m (RHU40H-130). Peak torque reaches up to 1,458 N.m on the largest model. Weight ranges from 382g (RHU14L without brake) to 6,150g (RHU40H with brake). Hollow bore diameters are 8mm, 10mm, 13mm, 18mm, and 22mm across the range.
While the RHU is optimized for humanoid robots, its large hollow bore and long service life make it suitable for any application requiring dense internal cable routing and extended operational cycles. Potential applications outside humanoid robots include collaborative robot arms with force feedback, medical rehabilitation robots, teleoperation platforms, and service robots that require force-controlled human interaction.
Most robot joints operate in position control mode — the controller commands a target angle, and the actuator moves to it. This works for repetitive industrial tasks where the environment is predictable. But humanoid robots operate in unstructured environments alongside people. A position-controlled arm that collides with a person will keep pushing until it reaches its target angle, regardless of the force applied. Force control solves this by making the joint aware of — and responsive to — the forces acting on it.
There are two main methods for estimating joint torque:
| Method | How It Works | Accuracy | Cost | Typical Use |
|---|---|---|---|---|
| Current-based (sensorless) | Estimates torque from motor phase current using KT calibration | ±5–10% | No additional hardware | Basic collision detection, cost-sensitive applications |
| Torque sensor (sensor-based) | Measures actual output torque via strain gauge or other sensor at the reducer output | ≤0.5% FS | Higher (sensor + signal conditioning) | Compliant manipulation, human-robot interaction, precision force tasks |
Current-based estimation is simpler and cheaper — the EYOU PH series with KT calibration achieves ±5% accuracy, sufficient for basic safety functions. But for applications requiring precise force modulation — like handing objects to a person, wiping a surface with controlled pressure, or performing physical therapy — a dedicated torque sensor is necessary. The RHU-F integrates this sensor at the harmonic reducer output, providing ≤0.5% FS accuracy at ≥5kHz sampling.
Force-sensing joints enable impedance control, where the joint behaves like a virtual spring-damper system. Two parameters define the behavior:
By adjusting K and D in software hundreds of times per second, a single physical actuator can behave like a rigid strut when supporting weight, then instantly become compliant to absorb an unexpected bump. This is how humanoid robots achieve natural-feeling physical interaction without dedicated mechanical compliance mechanisms.
In a typical humanoid robot, force control operates at multiple levels:
The joint-level torque loop is the foundation. If the torque sensor is slow, inaccurate, or introduces lag, the higher-level controllers cannot achieve stable force control. This is why sensor specifications like sampling frequency and accuracy matter at the actuator level — they set the ceiling for what the robot can achieve in force-controlled tasks.
These applications share a common requirement: the joint must sense real-time output torque accurately enough to modulate its behavior based on external forces. This is the core value proposition of the RHU-F force-control variant within the humanoid robot joint actuator ecosystem.
