Article · AI translation
On October 27, Xiaomi Technology disclosed the results of its exploration of bionic dexterous hand technology: taking flexible manufacturing scenarios at Xiaomi's car factory as the starting point for its design, it presented its latest technical thinking on a high-degree-of-freedom bionic dexterous hand.
Xiaomi said its robotics team collected complete data from 1827 standard operating procedures (SOPs) at the car factory and analyzed and validated it. Simulation ablation analysis based on real operating data showed that, in the 17–21-degree-of-freedom range, the average stable grasping success rate of the bionic dexterous hand was 15%–38% higher than in the 6–15-degree-of-freedom range, with 21 degrees of freedom delivering the greatest gains in factory operation coverage and grasping success rate. In this scenario, when the number of degrees of freedom was increased further to 23 or more, improvements in grasping ability slowed markedly, and the marginal returns from adding degrees of freedom declined.
Based on these findings, Xiaomi designed a new 21-degree-of-freedom bionic dexterous hand. It uses a hybrid drive and tendon-cable tension control system to achieve precise control of grasping force and improve reliability and maintainability, while also systematically optimizing heat dissipation and tactile sensing across the entire palm.
Chen Duan, director of the Center for Innovative Development of Digital Economy and Economic Integration at the Central University of Finance and Economics and deputy director of the Intelligent Products and Industry Working Committee of the Chinese Association for Artificial Intelligence, said that Xiaomi's R&D approach of “scenarios defining products,” which takes real factory operations as the starting point for design, is a more durable competitive moat than hardware specifications. “The tactile, force and position, and assembly sequence data collected in the workshop are scarce scenario-specific assets. They can be used to train the dexterous hand's operating strategies and can also continuously feed back into Xiaomi's entire AI system, improving its ability to transfer across scenarios.”
This article was provided by Xiaomi and republished by QbitAI with authorization. The views are those of the original author.