[Edaily Reporter Shin Young-bin] #SPG, a company specializing in actuators and precision gearheads for humanoid robots, has developed technology that enables robot joints to autonomously diagnose the type and location of a malfunction and, in the event of an anomaly, allows adjacent joints to distribute the force and bring the robot to a safe stop.

SPG announced on the 18th that it has developed “AI Fault Diagnosis and Active Fault-Tolerant Control” technology for humanoid robot joints and has filed a patent application.

This technology is designed to enhance joint stability, which is considered a key challenge in the commercialization of humanoid robots. Actuators, which serve as the joints of humanoid robots, are core components that account for a significant portion of the robot’s production cost. At the same time, they are also the parts most likely to reach their limits first due to heat generation, wear, and gear damage.

Previously, when a robot joint malfunctioned, engineers often had to disassemble the robot to identify the cause of the failure. SPG’s technology is distinguished by its ability to detect malfunctions using only internal signals generated during motor operation, without the need for separate vibration sensors or external diagnostic equipment.

SPG explained that by analyzing the signals generated when the motor is in motion, the system can distinguish between various types of failures—including electrical issues such as winding short circuits and magnet damage, as well as mechanical issues like bearing wear, gear breakage, and shaft misalignment. The company collected data by repeatedly operating test samples that simulated actual failures and trained an AI model on this data to improve diagnostic accuracy.

The company likened this technology to a human, explaining that “it works on the same principle as knowing whether an injury is a bruise or a blister.” This means that a robot’s joints can assess their own condition in real time and detect signs of failure early on.

Another key feature is “active fault-tolerant control.” By connecting dozens of joints via a single real-time communication network, the system immediately identifies which joint is malfunctioning when an anomaly occurs in a specific joint. Subsequently, if the force in that joint drops, adjacent healthy joints share the torque load to prevent the robot from suddenly collapsing.

This is expected to reduce secondary accidents—such as falls or dropped workpieces—in environments where humanoid robots work alongside humans. The company explains that this technology goes beyond simple fault detection to help the robot safely halt its movements or maintain its posture even after a failure occurs.

SPG independently mass-produces three types of precision reducers—planetary, harmonic, and RV—in South Korea. The fact that the company handles motors, reducers, and controllers as an integrated whole served as the foundation for this technological development. The company explained that since sensorless self-diagnosis requires access to internal motor control signals, the ability to comprehensively understand both the motor and the controller is crucial.

Professor Heo Jin of Incheon National University, an expert in motor fault diagnosis, and Professor Choi Dong-il of Myongji University, formerly of KAIST’s HUBO Lab, participated in the development of this AI fault diagnosis technology. Through collaboration with external researchers, SPG is advancing technologies that enhance the safety and reliability of humanoid actuators.

Following its patent application earlier this year for a structural design that reduces heat generation—a chronic issue with actuators—SPG has now added a patent for its fault diagnosis software technology. The company explained that through this approach, it is simultaneously addressing the two core challenges of humanoid joints by solving “heat generation with hardware and faults with software.”

SPG is currently preparing for mass production of its next-generation actuator, the “SDD (SPG Direct Drive),” which integrates a motor, reducer, and controller. The company is also jointly developing a Korean-standard humanoid actuator with the Korea Institute of Machinery and Materials.

SPG CEO Yeo Young-gil stated, “For robots, joints are ultimately their lifeblood,” adding, “While there is fierce competition to make robots smarter using AI, few companies have focused on enabling robots to diagnose their own problems and prepare for safety incidents.” He continued, “SPG will lay the foundation for robots that, like humans, are aware of their own condition and can work safely.”