Photo: Snapshot from CCTV News

 

The Beijing Innovation Center for Humanoid Robotics recently released the motion control algorithm framework of the Beijing half-marathon champion robot " Tien Kung," filling a gap in the open-source domain for high-performance humanoid robot motion control frameworks, CCTV News reported on Tuesday.

The algorithm framework "Tien Kung-Lab," which is independently developed by the Beijing Innovation Center for Humanoid Robotics, can train humanoid robots to autonomously adapt to complex terrains and move quickly in a short period of time. Its release is expected to provide core fundamental technical support for more humanoid robots to leave the laboratory and perform tasks in highly complex environments.

The center also opened several virtual robot training grounds that support full-scenario interaction, enabling large-scale embodied intelligence data collection and training.

According to CCTV News, because the number of robots that can operate simultaneously in real-life settings such as living rooms, lounges, kitchens, supermarkets, and factories is limited, developers have constructed digital twin simulation scenarios to achieve large-scale embodied intelligence data collection and training.

Six major digital twin robot training grounds developed by the Beijing Innovation Center for Humanoid Robotics and the Beijing Institute of Architectural Design have been made available to global developers. The simulation of over 200 types of high-precision movable objects, including drawers, cabinets, refrigerators, can be achieved in the training grounds. Developers can train directly based on existing environments or quickly generate new object assets to expand simulation scenarios.

Global Times