A collaborative effort among Tsinghua University, Peking University, Shanghai Jiao Tong University, Wuhan University, and several other institutions has resulted in the launch of the human - machine collaborative training framework known as MotionTrans. This innovative framework empowers robots to acquire new skills merely by observing human actions, eliminating the necessity for explicit demonstrations. This achievement represents a major leap forward in the realm of robot learning.

Traditionally, robot training has heavily depended on vast quantities of real - world demonstration data. Gathering such data is not only time - consuming but also expensive. In contrast, MotionTrans makes use of VR (Virtual Reality) devices to precisely capture the subtle hand movements of humans. Researchers employed portable VR devices to collect data and subsequently built a dataset comprising 3,213 demonstrations.

The core technology of MotionTrans revolves around converting human action data into a format that robots can comprehend. Additionally, it adjusts parameters such as speed and comfort zones to guarantee the safe and stable operation of robots.