Our system uses a simple RGB camera to generate 3D human skeletal information from 2D images. This allows the robot’s arms to instantly and easily mimic an operator’s arm postures. For leg imitation, the robot uses an end-to-end learning architecture. This architecture improves the accuracy of action recognition and demonstrates stronger generalization capabilities. In practical applications, we expect robots to perform tasks like picking and transporting goods in place of humans.

Our system uses a simple RGB camera to generate 3D human skeletal information from 2D images. This allows the robot’s arms to instantly and easily mimic an operator’s arm postures. For leg imitation, the robot uses an end-to-end learning architecture. This architecture improves the accuracy of action recognition and demonstrates stronger generalization capabilities.

This technology has the potential to improve operational efficiency, reduce human labor in heavy or repetitive tasks, and reduce risks in hazardous environments. In addition to logistics, the system’s ability to accurately mimic complex human movements also open the door to applications in manufacturing, healthcare, services, and even household assistance.