Chinese company's humanoid robot wins two gold medals in world competition

The second World Humanoid Robot Sports Meet kicked off today, with the Chinese company Zhuyun showcasing its core products, including the G2, X2, and A3 models. In the completed 'Scenario' events, the G2 model secured two gold medals, three silver medals, and two bronze medals, placing Zhuyun at the top of the medal table. The competition introduced a new 'Scenario' category, simulating real-world environments across industries such as industrial, hotel, home, and logistics. The G2 model won gold in the 'Fire Emergency Scenario' and 'Library Scenario' events. In the 'Fire Emergency Scenario,' teams had to complete three core tasks: identifying hazardous materials, shutting off abnormal valves, and using a fire extinguisher to put out a fire. The fire extinguisher weighed between 4.5 to 5 kilograms, far exceeding the load capacity of most humanoid robot end-effectors. Zhuyun opted for a structural addition on the wrist rather than a custom gripper, using the wrist's structural strength to lift the extinguisher and place it in a dedicated bucket with a 20-degree tilt. In the 'Library Scenario,' Zhuyun's team, in collaboration with Tsinghua University and Shanghai Jiao Tong University, competed in three stages: book retrieval and transport, shelving, and correcting misplacements. Any failure in these stages could interrupt the task, requiring high robustness across the entire process. The team created a high-precision map with key points, used localization and navigation algorithms to calculate the optimal path, and implemented an open-source visual model to recognize book spine labels, building a stable and reliable autonomous decision-making system for the robot. According to official information, the G2 is a mature embodied intelligent product already deployed in numerous real-world scenarios. It features a 5-kilogram force-controlled harmonic joint arm capable of handling most objects, and a domain controller with 2070 TOPS computing power, enabling efficient algorithm execution on-site and eliminating reliance on external computing power and networks, ensuring more stable performance.