Summary of Key Points
Currently, commercially available robots generally use low-cost two-finger grippers, which have the problem of being clumsy and inefficient (for example, they struggle to hold apples securely or crush薯 chips, and their movements are shaky). More flexible, high-degree-of-freedom dexterous hands have not been widely adopted due to high costs (ranging from hundreds of thousands of yuan per unit), immature technology (lack of drive models and significant data gaps), and an incomplete supply chain. The industry is working to overcome the "impossible triangle" of performance, reliability, and cost by optimizing technical approaches (linkage/rope drives/prime direct drives), developing custom core components, and advancing large-scale dexterous operation models. There is substantial potential for market growth in the future (with sales expected to reach 70,000 units by 2026), but this will require breakthroughs in technology similar to those of GPT3.0.
I. The Common Issues with Gripper Robots
Most robots used in stores and homes today employ two-finger grippers, which have several significant drawbacks:
- Inability to handle soft or deformed objects securely: Cleaning robots using these grippers often drop apples (since the training uses artificial objects that differ from real apples), and "Gebert" robots in convenience stores may crush薯 chip bags due to poor force control.
- Unsmooth movements: Robots pause when switching tasks, and their robotic arms may shake suddenly (this is not a hardware issue but rather a result of model decisions based on training data that do not reflect real-world scenarios).
- Limited functionality: Cleaning robots can only perform basic tasks like organizing tables and folding clothes, and even some services have been discontinued (such as the 58-to-Home robot cleaning service, which is no longer available for reservation through a mobile app).
Why are gripper robots so unreliable? Their design is simple and cost-effective, but they lack advanced control algorithms. With only two contact points, they struggle to maintain balance when objects deform, and their range of motion is much limited compared to the 21 degrees of freedom in a human hand.
II. Why aren’t Dexterous Hands More Popular?
Dexterous hands offer the same level of flexibility as human hands (with over 20 degrees of freedom), but they have not become widespread:
- High cost: These robots can cost hundreds of thousands of yuan, with models featuring 6–10 or more degrees of freedom costing several thousand to over a hundred thousand yuan. The main reason is the high cost of drive units (motors and gearboxes), as well as an immature upstream supply chain (customized micro-motors are expensive to procure).
- Lack of drive models: Current AI models can only control grippers; they are not yet capable of enabling dexterous hands to perform complex tasks like opening bottles or tying shoelaces, which represents a major challenge for the industry.
- Data gaps: Collecting data on how dexterous hands interact with the real world is costly (e.g., measuring the force and angle required to grasp different objects), and there is insufficient data available for development.
III. The Competition of Drive Technologies for Dexterous Hands
There are three main drive methods, each with its advantages and disadvantages:
- Linkage drives: Similar to robotic arms, these use links to transmit force, allowing them to handle heavy loads but limiting flexibility and natural movement.
- Rope drives: Simulate human tendons using ropes, providing a lightweight and cushioned mechanism that reduces the risk of damaging objects; however, ropes are prone to wear and tear.
- Prime direct drives: Motors directly control the joints, offering precise control and fast response times but at a higher cost and larger size (which can lead to overheating).
Companies are choosing the drive method that best suits their needs. For example, Wujit Technology uses prime direct drives, while LingqiaoZhiNeng employs rope drives, each with its own focus.
IV. Overcoming the “Impossible Triangle”
To resolve the contradiction between high performance, reliability, and low cost, companies are taking various approaches:
- Design improvements: Wujit Technology has integrated 20 motors within the finger joints to improve heat dissipation, allowing for longer operation times; LingqiaoZhiNeng has placed some motors in the palm area to reduce algorithm complexity and enhance response speed.
- Customized components: Companies like Wujit Technology design their own components and outsource manufacturing, reducing costs compared to similar products. LingqiaoZhiNeng’s self-developed drive modules have reduced the cost of its second-generation products by 50%.
- Advancing large-scale models: Companies are collecting data on hand movements (e.g., grasping techniques) and collaborating with embodied intelligence companies to train more advanced models, hoping for breakthroughs in the near future.
V. A Promising Future?
According to GaoGong Robotics, sales of dexterous hands in China are expected to increase by 236% to 19,200 units in 2025 and further rise by 265% to 70,200 units in 2026. However, the industry agrees that several factors are necessary for widespread adoption:
- Breakthroughs in embodied intelligence models (similar to GPT3.0, which can enable dexterous hands to perform complex tasks).
- A mature supply chain with mass-produced components and reduced costs.
- The establishment of practical use cases (e.g., in humanoid robots and manufacturing).
Companies are working hard to make dexterous hands more reliable and affordable, aiming to integrate them into everyday life.
This news article highlights a simple principle: for robots to function effectively like humans, their "hands" are crucial. Currently, either the grippers are inefficient or the dexterous hands are too expensive. The industry is striving to make intelligent and cost-effective hands a reality. Let’s wait and see how things develop!