What will the world be like when robots have dexterous hands?

"The real challenge in robot R & D lies in hardware design, especially the complex engineering problems of hands and forearms."

This seemingly simple judgment precisely points out a crucial bottleneck in the development of humanoid robots to date - having a pair of "hands" that can operate as flexibly as human hands is far more difficult than teaching robots to walk or run.

It's true that it's not easy to make a robot walk stably. However, for it to truly enter real - life scenarios and complete tasks such as serving tea, operating tools, and fine assembly, a dexterous hand becomes an indispensable key. This is not only a technical challenge but also a dividing line for whether robots can move from demonstration to practical use.

What's worth looking forward to is that with breakthroughs in materials, sensing, and control technologies, dexterous hands capable of threading a needle, holding an egg, playing the guzheng, and making coffee made surprising appearances at the recent WAIC (World Artificial Intelligence Conference) and WRC (World Robot Contest) and are accelerating their way into daily life.

Today, let's explore the innovative world of dexterous hands and see how these growing "hands" will change our lives and extend the boundaries of human capabilities.

How are dexterous hands "forged"?

The hand is one of the greatest advantages evolution has given to humans. It consists of 27 bones, 29 joints, and hundreds of muscles and ligaments working in coordination. It can perform delicate actions like picking up an embroidery needle and can also carry heavy objects of dozens of kilograms.

When we expect robots to truly enter human daily life, we also need a pair of dexterous "hands". Legs allow robots to walk and jump, but only hands can enable them to truly operate tools and interact better with the environment. It can be said that the dexterous hand is the "last mile" for humanoid robots to truly be implemented and serve humans. Data shows that in the long run, dexterous hands are expected to account for 20% - 30% of the total cost of humanoid robots, second only to body execution, and are one of the most important hardware components.

The most core part of a dexterous hand lies in its mechanical and structural design, which can be divided into structural form, driving mode, transmission mode, sensing mode, and materials and other aspects. Currently, there is no unified technical standard in the dexterous hand industry, and multiple technical routes co - exist. There is no conclusion as to which technical route has more advantages. This "contention of a hundred schools of thought" is particularly evident in the three most critical components - driving mode, transmission mode, and sensors.

Data source: Shadow Robot official website, Robotics and Computer - integrated Manufacturing, Soochow Securities Research Institute

Driving mode: How to train the "muscles"?

The driving system is the "muscle" of a dexterous hand, equivalent to the muscle tissue in the human body that provides power for the hand. It directly determines the performance ceiling and practical application potential of the dexterous hand. Currently, the mainstream driving modes mainly include four types: motor - driven, hydraulic - driven, pneumatic - driven, and shape - memory alloy - driven. Among them, motor - driven has become the most mainstream choice in industrial applications due to its comprehensive advantages such as small size, fast response, convenient control, high stability, and excellent precision.

Three commonly used motors for dexterous hands:

Coreless motor: It has an ultra - lightweight design. The latest diameter can reach 3 - 4mm, and it has a fast rotation speed. The energy utilization rate is generally between 75% - 90%. It can better meet the requirements of dexterous hands for high - power density and high - dynamic response, and is the main choice for the current electric - drive scheme of dexterous hands.

Brushless DC motor: It retains the complete motor structure, and its performance emphasizes control accuracy. It is mostly used in general industrial automation.

Frameless torque motor: It has a special design, only retaining the core components, and pursues high - torque density, suitable for high - integration robots.

Transmission mode: How to connect the "tendons and bones"?

The transmission system is the "ligaments and bones" in a dexterous hand. It is not only responsible for transmitting power but also plays a role in supporting the structure. As dexterous hands develop towards high - load - bearing and high - precision directions, transmission technology is undergoing a paradigm shift from a single mode to a composite system. The transmission schemes of dexterous hands include link - transmission, tendon - rope transmission, gear/worm - gear transmission, etc. Currently, the most mainstream methods are the tendon - rope transmission scheme and the composite transmission scheme of "tendon - rope +".

In fact, most early multi - fingered dexterous hands adopted the link - transmission scheme. This scheme uses a combination of series and parallel connections of multiple link structures. The motor is combined with a reduction device for deceleration, and then the link structure is used to transmit power to the fingers. The advantage is high stiffness, which can grasp large objects and has a compact structural design. The disadvantage is that the transmission structure is very complex, with a low degree of anthropomorphism and insufficient flexibility.

In order to achieve greater flexibility and save more space, the tendon - rope transmission scheme emerged. To a certain extent, it simulates the tendon structure of the human hand. While improving the grasping speed, it can also achieve flexible control similar to that of the human hand, making it suitable for transmission occasions with limited space and a large number of required driving degrees of freedom.

Lingxin Qiaoshou is the only domestic company capable of commercially mass - producing dexterous hands with both link structures and tendon - rope structures. Its research - version Linker Hand uses the "tendon - rope + link" scheme, with 42 degrees of freedom. Each finger can independently achieve 9 degrees of freedom of motion control and is equipped with a fully - driven design that can rotate 360 degrees freely. It can not only complete high - difficulty actions such as spinning a pen and playing with walnuts but also has a load - bearing capacity of 5kg, exceeding the limit of human fingers.

Sensors: Where does the "tactile sense" come from?

Sensors are the "skin" and "nervous system" of a dexterous hand. They not only endow the robot with the ability to sense the external environment but also provide key information support for its delicate operations.

The sensors used in dexterous hands can be divided into two major categories: internal and external. Internal sensors are mainly used to monitor the state of the hand itself. For example, motion sensors and force sensors can obtain real - time joint angles, positions, and motion dynamics to ensure the stability and accuracy of actions. External sensors include proximity sensors and tactile sensors, which are used to identify the position, shape, and force - bearing situation of the target object, enabling the dexterous hand to adaptively complete complex tasks such as grasping and manipulation. The two types of sensors work together to form the perceptual basis for the dexterous hand to interact with the physical world.

Currently, tactile sensors have become a standard configuration for dexterous hands, but there are different focuses in technical implementation. For example, the Pacini DexH13 innovatively integrates multi - dimensional tactile and AI vision dual - modality and is equipped with 1140 ITPU tactile sensing units to improve the perception accuracy in complex scenarios. The dexterous hand of Zhiyuan integrates MEMS - based tactile sensors and vision sensors, which can sense the shape, material, and even temperature of objects.

Picture source: Screenshot from the video account of Zhiyuan Robot

Three types of players, three ways to break the situation

Although the technical routes of dexterous hands have not yet converged, the market is getting more and more lively. Data from the Zhongshang Industry Research Institute shows that the global market size of robot dexterous hands was $1.7 billion in 2024 and is expected to exceed $3 billion by 2030.

Facing the multi - billion - dollar market, various players are continuously pouring in like mushrooms after rain. They are looking for suitable implementation scenarios based on their own technical strengths and market directions. These players can be roughly divided into three categories:

The first type is the "self - developed ontology school", that is, downstream complete - machine manufacturers expand upstream. Representative enterprises include Unitree, Zhiyuan, Magic Atom, Xingdong Jiyuan, Ubtech, Fourier, Zhongqing, Lingchu Intelligence, etc. These players mainly focus on the R & D of humanoid robot complete machines and develop dexterous hands as core components of the complete - machine system simultaneously. They emphasize the deep coupling of dexterous hands with their own robots in motion control and perception fusion to achieve more efficient and precise collaborative operations.

For example, the XHAND1 of Xingdong Jiyuan can handle multiple tasks on its Q5 and L7 robots. In April this year, Unitree Technology launched the Unitree Dex5, which has 20 degrees of freedom in a single hand, supports smooth reverse driving, and is equipped with 94 sensitive contact points. From Unitree's demonstrations, it can be seen that the humanoid robot H1 equipped with Dex5 can easily perform high - precision actions such as playing poker, solving a Rubik's cube, and turning pages of a book.

The second type is the "vertical new forces". These players focus on the R & D and production of single dexterous - hand products. They launch targeted products according to the needs of segmented markets and pay attention to product technological innovation and scenario adaptability. Representative enterprises include Lingxin Qiaoshou, Lingqiao Intelligence, Zhongke Guiji, etc. These vertical manufacturers have launched many industry - benchmark products, and some enterprises have formed certain technical barriers and market competitiveness in segmented fields.

For example, the high - degree - of - freedom version of the Linker Hand series of Lingxin Qiaoshou focuses on the scientific research and education field, providing advanced scientific research tools for universities and research institutions for research in fields such as robot motion control and perception algorithms. The low - degree - of - freedom version has expanded to lightweight fields such as beauty and health care.

The third type of players is the "cross - border entrants", that is, upstream component manufacturers expand downstream. Relying on their technical accumulation and production capabilities in the field of core components, they horizontally enter the dexterous - hand market and provide cost - effective dexterous - hand products or core modules for complete - machine manufacturers. Representative enterprises include Zhaowei Machinery & Electronics, RoboSense, Leadshine Technology, etc.

For example, RoboSense, a leading player in the lidar field, released its first - generation dexterous hand as early as 2024 and officially launched its second - generation dexterous hand Papert2.0 in early 2025. Leadshine Technology, which has strong technical strength in fields such as stepper motors and servo motors, combines motor technology with dexterous - hand design and has launched dexterous - hand products suitable for scenarios such as industrial grasping and handling.

The continuous competition among new and old players is constantly promoting the technological iteration of the dexterous - hand industry and accelerating the industrialization process of dexterous hands. Overall, there are several obvious trends:

Firstly, the degree of freedom is continuously increasing. "Being most like a human hand" has always been the long - term goal of dexterous - hand R & D. Early dexterous hands usually had about 6 degrees of freedom. At present, the mainstream products have generally increased to 12 - 20 degrees of freedom, getting closer and closer to the number of degrees of freedom of a human hand. Even the research - version Linker Hand L30 of Lingxin Qiaoshou is equipped with 42 degrees of freedom, far more than that of a real human hand. However, an increase in the degree of freedom also means an increase in control difficulty. Enterprises no longer simply pursue quantity but pay more and more attention to the effective allocation and collaborative control of degrees of freedom, enabling each finger to move independently and cooperate tacitly.

Secondly, lightweight and miniaturization have become the main innovation directions for domestic enterprises. Early international dexterous - hand products, such as the Shadow Dexterous Hand (standard version), were nearly twice the size of an adult male's palm and were more suitable for heavy - duty robotic arms. With the popularity of humanoid robots, there is a greater need for lightweight and highly flexible dexterous hands. Enterprises are also making targeted breakthroughs. For example, Unitree's Dex5 - 1P, Lingxin Qiaoshou's Linker Hand O6, Lingqiao Intelligence's DexHand 021, and Leadshine Technology's DH2015 have all controlled their weights to within 1kg.

Picture source: Screenshot from the video account of Lingxin Qiaoshou

Thirdly, intelligence has become a difficult problem that dexterous hands have to overcome. Currently, dexterous hands are relatively lacking in visual data, precise behavior data, and tactile data. There is still a long way to go to achieve true large - model - driven operation. Of course, enterprises are also constantly making efforts in intelligence. For example, the training mode of the dexterous hand of Dahan Robotics uses reinforcement learning based on data such as tactile, joint position, and motor position, which helps to train a usable control strategy in a short time. Pacini has officially launched a data - collection factory, and the collected data will be used for self - training or sold. Zibianliang Robotics has achieved autonomous operation of a 20 - degree - of - freedom dexterous hand with its self - developed end - to - end large model.

It can be predicted that the next few years will be a critical stage for dexterous hands to move from "technical verification" to "scenario in - depth exploration".

In these scenarios, dexterous hands are "showing their prowess"

Meanwhile, the application scenarios of dexterous hands are constantly expanding from traditional fields to emerging fields, setting off a series of "operation waves" in industries such as industrial manufacturing, medical rehabilitation, and life services.

In the field of industrial manufacturing, dexterous hands break the limitations of the "rigid production" of traditional robotic arms and enable the production line to truly achieve the upgrade of "flexible intelligent manufacturing". For example, many new - energy vehicle enterprises apply dexterous hands to battery production lines. The assembly of new - energy batteries requires "gentle handling and precise stacking". Dexterous hands can stack battery cells into groups with extremely high precision, and the pressure error of each battery cell does not exceed 2 Newtons, effectively avoiding safety hazards caused by battery - cell deformation.

In the field of medical rehabilitation, dexterous hands become the "third hand" of doctors, bringing hope to patients in three directions: surgery, rehabilitation, and prosthetics. For example, a medical dexterous hand equipped with multi - dimensional tactile sensors can ensure the surgical precision reaches 0.1 mm, even higher than the precision achieved by a doctor holding an instrument personally. A rehabilitation robot equipped with a dexterous hand can provide personalized rehabilitation training programs for rehabilitation patients, helping them gradually recover nerve function and hand - movement ability. A bionic prosthetic limb based on a high - degree - of - freedom dexterous hand can not only perform delicate operations such as unscrewing a bottle cap and tying shoelaces but also feed back signals through tactile sensors, allowing patients to truly feel "tactile sense".

In the field of life services, dexterous hands are entering daily life and becoming considerate "housework assistants" and "service helpers". For example, in the application of home - based elderly care, through the high - precision operation of dexterous hands, they can assist the elderly in dressing, feeding, taking medicine, etc., improving the quality of life of the elderly. In restaurants