Xiaomi unveils a bionic hand with 21 degrees of freedom as competition in dexterous hands shifts toward software and manipulation data

What has changed in Xiaomi’s new bionic hand?

The new-generation bionic dexterous hand unveiled by Xiaomi was designed around flexible manufacturing scenarios at its automotive factory. It uses a hybrid drive with 21 degrees of freedom and 26 motors, and has tactile sensing across the full palm at 1500 points.
In grasping tests across multiple poses, the new hand achieved both more human-like grasping and a higher success rate than the previous version with fewer degrees of freedom.

The report finds that the degrees of freedom, load capacity and transmission methods of mainstream dexterous hands are gradually converging, making it increasingly difficult to differentiate products simply by adding degrees of freedom or load capacity.
Domestic production of miniature motors, precision transmission components and sensors has lowered the barriers and costs of hardware manufacturing. When manufacturers use similar component configurations, their ability to perform manipulations in real-world scenarios depends more heavily on algorithms and manipulation data.
Hong Jinping of Huachuang Securities believes that data, simulation environments, grasp-planning algorithms and hand-eye coordination models will become core barriers to entry for dexterous-hand companies, while software capabilities and data assets will also become more important to valuations.

Lingxin Qiaoshou has built DexSkill-Net through real-world simulation learning and accumulated more than 200T of hand data. Lingqiao Intelligence’s DexCanvas brings together more than 1000 hours of real human multimodal demonstration data and 100 thousand hours of synthetic data from physical simulations.
These datasets support the development of dexterous manipulation capabilities and show that some manufacturers are investing resources in software and data instead of only adding hardware features.

Changying Precision’s products include high-precision components such as transmission parts for dexterous hands, and it provides Tianji Robot with core components and complete-machine assembly and testing services. Its motion-control and movement-execution functions connect to the customer’s system for perception, understanding and planning, forming a complete embodied intelligence system.
OPT applies industrial vision technology to robotics and is developing visual perception components such as cameras and lidar to support needs including environmental perception, positioning and navigation.