Industrial physical artificial intelligence (AI) firm Diden Robotics and human motion AI startup Movin are joining forces to create a dataset that will integrate authentic human movements into industrial humanoids.
The collaboration aims to boost robot motion learning by combining Light Detection and Ranging (LiDAR)-based motion capture, multimodal sensor data, and industrial humanoid development expertise.
Industry insiders report that deploying humanoid robots in human-occupied spaces requires defining how people navigate complex environments. This involves capturing precise data on foot placement angles on stairs and ramps, body contortions in narrow passages, and hand-eye coordination when handling objects at workstations. Such detailed data is crucial for simultaneous control of walking, balance, posture, and manipulation.
Data is the linchpin of humanoid performance. To effectively train core behavioral models for walking, balance, full-body control, and manipulation, it’s essential to incorporate precise three-dimensional (3D) motion and spatial data reflecting human movement, balance maintenance, and object interaction in complex industrial and urban settings.
Movin’s technology uses multiple LiDAR cameras to simultaneously capture full-body movements and surrounding spaces, reconstructing this information into 3D human and environmental models. The company recently partnered with NVIDIA to provide real-time motion capture technology for their AI human model SOMA, contributing to the global AI human ecosystem.
In this partnership, Movin will adapt its motion and environmental data to match Diden Robotics’ humanoid joint structure, length, and range of motion, creating movements that robots can realistically replicate.
Diden Robotics, a startup launched in March 2024 by four Korea Advanced Institute of Science and Technology (KAIST) mechanical engineering and humanoid robotics researchers, has secured contracts with Korean shipbuilders for their Diden Spider. This robot performs welding, inspection, and maintenance tasks while navigating steel walls and ceilings. The company is currently in talks for additional contracts.
Leveraging Movin’s data, Diden Robotics is accelerating its humanoid development. The company reports achieving a humanoid standing 175 cm (about 5 feet 9 inches) tall, weighing 90 kg (about 198 pounds), with a continuous payload capacity of 30 kg (about 66 pounds). They plan to showcase the robot’s capabilities at RoboWorld this November.
Diden Robotics Chief Executive Officer (CEO) Kim Jun-ha emphasized that enhancing humanoid performance requires high-quality multimodal data capturing both human movements and spatial context. The partnership with Movin will help them secure robot-optimized data and strengthen the learning systems for immediate application in robot development.
Movin CEO Choi Byeol added that the ability to precisely capture human movements and adapt them to a robot’s physical structure is fundamental to humanoid robot learning.