J-Mex Unveils Human-to-Robot Motion Conversion Tech at Computex 2026

Taiwanese company J-Mex brought its new AgileMaster platform to Computex 2026, designed for remote control of humanoid robots and AI data collection. The development targets the rapidly growing market for Physical AI, digital twins, and human-machine interaction.

At the core of the solution is the V100-R Plus industrial motion capture system using inertial measurement unit (IMU) sensors, which tracks the operator’s movements and transmits them to the robot in near real time. The company claims the platform’s architecture delivers millisecond-level synchronization, enabling AgileMaster to be used for robot teleoperation, learning from human demonstration, and creating digital twins. A key feature is Motion Retargeting technology, which converts human motions into commands for various robot types, including humanoid models and robotic manipulators. The developers have also baked in support for popular robotics tools like ROS2, URDF, and the NVIDIA Isaac Sim ecosystem, widely used for simulation and training of robotic systems.

According to company reps, AgileMaster’s main goal is to generate high-quality human motion datasets that can then be used to train cutting-edge robotics models. Such models are seen as a promising frontier in humanoid robot development and Physical AI. During a chat with VGTimes, a J-Mex rep also stressed the importance of working with NVIDIA on this front.

It’s worth noting that shortly before the show, AgileMaster snagged a Best Choice Award 2026 in the Robotics & Drones category, handed out annually to the most exciting innovations.

Beyond the AgileMaster platform itself, the company is showcasing its own motion capture technologies, already being used in projects related to training humanoid robots and collecting data for AI systems. J-Mex’s solutions have previously been featured in robotics demos at NVIDIA events and with other industry partners. Meanwhile, robots are already beating humans at running.

What do you think — will teleoperation systems and human-motion learning speed up the arrival of genuinely useful humanoid robots, or is the industry still a long way from mass adoption of such machines? Drop your thoughts in the comments.