Beijing, August 26, 2026 — The second World Humanoid Robot Games has pushed a critical question to the center of the global robotics stage: how can humanoid robots move from spectacular demonstrations to useful work in factories, homes, hospitals, and public spaces? The answer increasingly points to the hand. While legs allow humanoid robots to run, walk, balance, and enter human environments, hands determine what those humanoid robots can actually do once they arrive. The dexterous hand, a compact but extraordinarily complex module, has become one of the clearest measures of whether humanoid robots can become practical partners in real economic and social activity.
The scale of the event reflected how quickly the field is moving. The organizing committee office deputy director and Beiao Group chairman Zhao Dongwei said the second World Humanoid Robot Games expanded from 26 events to 51 events, from 280 teams to 666 teams, from about 500 robots to 2,056 robots, and from a three-day schedule to a five-day schedule. Beiao Group, a State-owned enterprise linked to the Double Olympic legacy, again jointly operated the event. For humanoid robots, the competition was not only a showcase of motion control or athletic ability. It was a stress test of perception, touch, force control, planning, and decision-making. The newly added dexterous hand special competition made that shift visible through tasks such as powder weighing, bottle cap prying, and picking up beans with tweezers.
These are ordinary actions for a person. For humanoid robots, they are a mountain at the fingertips. A human can weigh powder, open a bottle, or pick up a small object without consciously calculating friction, angle, pressure, or timing. Humanoid robots must integrate vision, touch, force control, and decision-making in real time. Every small movement becomes a test of hardware and software working together. The dexterous hand is therefore not an accessory. It is a gateway technology for humanoid robots that are expected to work in unstructured environments.
-
From Athletic Spectacle to Skilled Work: Why Hands Matter for Humanoid Robots
The dexterous hand was singled out for a dedicated competition because it defines the boundary between a machine that can perform and a machine that can work. Humanoid robots may impress audiences by running, jumping, or maintaining balance, but their economic value depends on whether they can manipulate objects safely, reliably, and intelligently. Legs decide where humanoid robots can go. Hands decide what humanoid robots can become. This distinction explains why the event organizers placed such emphasis on tasks that resemble real-world labor rather than staged demonstrations.
The competition design reflected that philosophy. In the building block challenge, humanoid robots were required to assemble blocks within a limited time and keep the structure stable for five seconds. This demanded close-range vision, complete planning logic, stable grip force, and precise control. In the electric tool assembly task, humanoid robots needed compliant force control and accurate positioning. In the tweezers bean picking task, they needed micro-force manipulation. In the cable connection task, they needed hand-eye coordination. The eight sub-events amounted to a demanding examination of dexterous hand capabilities, covering many of the skills that humanoid robots must master before they can be trusted with productive tasks.
Lingxin Qiaoshou, an embodied intelligence company, entered all eight dexterous hand competition events. The company also won the overall championship at the ninth Zhongguancun International Frontier Technology Competition with a project described as an embodied intelligence platform centered on dexterous hands and cloud intelligence. Through innovation in finger joint modules, self-developed micro motor technology, and core material breakthroughs, the company achieved autonomous control over key components. It addressed critical technical challenges in humanoid robot research and development and became a leading global supplier in the dexterous hand field. According to the company, eight out of every ten high-degree-of-freedom dexterous hands worldwide come from Lingxin Qiaoshou. That figure alone shows how concentrated and strategically important the dexterous hand supply chain has become for humanoid robots.
In the powder weighing event, humanoid robots had to use a spoon to take and place 20 grams of powder, with an error tolerance of plus or minus 0.5 grams. In the bottle cap prying event, humanoid robots had to use an opener to pry a cap from a glass bottle. The number of valid bottle openings completed within five minutes determined the score. This task required the dexterous hand to apply precise force in a narrow space, placing extremely high demands on force control accuracy, dynamic coordination, and real-time feedback. Zuo Jiaping, co-founder of Lingxin Qiaoshou, explained that it was not merely a matter of one hand holding the bottle steadily. The other hand also had to pick up the opener, and the posture, force, and position used to grasp the opener all affected the speed and quality of opening. The humanoid robots therefore had to truly use tools in a human-like way.
The pitch-pot event added another dimension. The pot had an inner diameter of 13 centimeters, and its center point was 1.5 meters horizontally from the throwing line. Humanoid robots had to maintain a standing posture for three minutes and complete the full sequence of grasping and throwing. The core test involved hand-eye coordination, visual target recognition, upper-limb trajectory planning, stable grasping by the dexterous hand, and precise control of throwing force and angle. It also examined static standing balance. Zuo Jiaping noted that the pitch-pot task required humanoid robots to see clearly, judge distance from the surrounding environment, hold the arrow accurately and steadily, and throw with precision. The robots needed explosive force, angle calculation, and a smart brain capable of estimating distance, force, and position. It was a combined examination of hand and brain for humanoid robots.
The performance of teams in the dexterous hand events was notable. Wang Shilong, chief judge of the dexterous hand special competition, said fully autonomous teams performed better than previously expected, and many achieved strong results. Because the fully autonomous mode is more difficult, such humanoid robots received scoring advantages over teleoperated modes, encouraging teams to develop robots capable of completing tasks independently. At the same time, even small changes in the material, hardness, or placement angle of an object could cause autonomous decision-making to deviate. This uncertainty in the real physical world remains one of the most important technical challenges, and it is precisely the direction the competition organizers hope to push forward.
Gong Xiao, a committee member and deputy director of the China Software Testing Center, said the competition is not the end but the starting point. Behind every competition task is a question drawn from real production and daily life. Testing through actual practice and using demand as a guide is the original mission of the event. For humanoid robots, that mission means moving from controlled arenas to messy, unpredictable environments where objects shift, surfaces vary, and human needs are not neatly scripted.
-
The World Robot Conference 2026: Dexterous Hands Move to Center Stage
Changes in competition design often mirror changes in industrial direction. That shift was already evident in 2026. On August 19, the 2026 World Robot Conference opened under the theme “Human-Machine Symbiosis, Production-Demand Integration.” Dexterous hands became a central attraction. As a core component of humanoid robots, the dexterous hand integrates mechanical structure, motors, sensors, and control algorithms. It is one of the most difficult core modules in the entire machine. This is not an exaggeration. The human hand has more than 20 degrees of freedom. A robotic dexterous hand borrows from that structural logic. The higher the degree of freedom, the richer the actions it can complete and the more work it can handle.
Yet the hand is also a space-constrained engineering problem. A robotic hand must fit dozens of joints, drivers, and sensors into an extremely small volume while maintaining response speed and reliability. The engineering difficulty is far greater than the simple rotation of large joints such as the upper arm or thigh. For humanoid robots, the dexterous hand is where mechanical design, microelectronics, sensing, and control algorithms collide. It is also where the promise of humanoid robots becomes tangible. A humanoid robot without capable hands can walk and gesture. A humanoid robot with capable hands can assemble, inspect, serve, assist, and adapt.
At the World Robot Conference, Lingxin Qiaoshou displayed multiple products and scenario application solutions, including a fully automatic robotic arm production line, the Linker Hand series of dexterous hands, and a robot band. The fully automatic robotic arm production line was composed of multiple Lingxin Qiaoshou workstations operating as a cluster. Each workstation adopted a dual-arm, dual-hand configuration equipped with Linker Hand series dexterous hands. Through visual recognition and force control technology, combined with anti-interference capability and generalizable tool use, the production line achieved multi-workstation collaborative operation. It could complete the full process from parts grasping and assembly to finished product off-line. Each workstation had the ability to operate tools, switch production skills, and coordinate processes. A single workstation could assemble components for the shoulder, elbow, upper and lower arms, wrist, and other parts of a robotic arm.
At the Zhaowei dexterous hand booth, flexible sensing materials covered fingertips, joints, and palms. The hand was linked to a tablet screen. When visitors pressed a fingertip, the screen displayed the force point, force magnitude, duration, and pressure change curve in real time. The company explained that the dexterous hand used a piezoresistive flexible array tactile solution, with hundreds of sensing points integrated into each finger. After identifying the magnitude and direction of force, the sensors returned data, and the system dynamically adjusted gripping force to prevent objects from being crushed or slipping. This kind of tactile feedback is essential for safe, flexible operation by humanoid robots in homes, hospitals, warehouses, and factories.
In 2026, domestic dexterous hands saw a concentrated wave of technical breakthroughs. Multiple indicators reached or surpassed comparable international products. Among the fully direct-drive dexterous hands from the embodied intelligence company Star Dynamics Era, the XHAND 1 has 12 active degrees of freedom, a maximum single-hand grip force of 80 newtons, the ability to lift 25 kilograms, and a repeated opening and closing speed of about 10 times per second. The company states that it can complete more than 100 operating tasks. The XHAND 1 PRO is configured with 21 active degrees of freedom and 18 tactile sensors, with repeat positioning accuracy of plus or minus 0.1 millimeters and backlash of plus or minus 1 millimeter. It is compatible with data collection gloves such as Manus Pro and Manus Haptic Pro, supports robotic arm interfaces such as UR and Franka, and covers communication protocols including RS-485, EtherCAT, and CAN.
These specifications matter because humanoid robots need hands that are not only strong but also gentle, not only fast but also precise, and not only repeatable but also adaptable. A hand that can lift 25 kilograms may also need to pick up an egg. A hand that can open and close 10 times per second may also need to pause, sense, and adjust. The dexterous hand is therefore a convergence point for mechanical power, tactile intelligence, and real-time control. It is also a major cost and reliability factor in the commercialization of humanoid robots.
Core components required for dexterous hands have now achieved domestic mass production and autonomous control. Scaled production lines have also reduced product costs, laying a foundation for the wider adoption of humanoid robots. This means the dexterous hand, often described as the “last centimeter” that allows robots to enter daily life and become true assistants, is moving from exhibition booths toward more concrete scenarios. For humanoid robots, the last centimeter is not a minor detail. It is the final link between intelligence and action, between computation and physical usefulness.
-
Capital, Production Lines, and the Industrial Ecosystem Behind Humanoid Robots
At the end of 2023, Beijing established two funds at the level of 10 billion yuan each: the Robot Industry Development Investment Fund and the Artificial Intelligence Industry Investment Fund. Both are managed by a professional institution under Beijing State-owned Capital Operation and Management, a municipal enterprise. This was only part of a broader effort. Beijing has established a government investment fund with a total scale of 100 billion yuan to increase investment in future industries represented by humanoid robots. As the core management platform for Beijing municipal government investment funds, Beijing State-owned Capital Operation and Management currently manages eight municipal government investment funds, building a capital matrix that covers strategic emerging industries and future industries.
In 2025, Beijing State-owned Capital Operation and Management participated in multiple investments in the embodied intelligence track, mainly focusing on leading Beijing-based humanoid robot companies. Lingxin Qiaoshou, which holds more than 80 percent of the global dexterous hand market, began its financing process in 2025 and completed six rounds within one year. Its shareholders include Sequoia China, Ant Group, CICC Capital, Beijing State-owned Capital Operation and Management, and other well-known institutions. The support from Beijing State-owned Capital Operation and Management helped core component companies such as dexterous hand developers rapidly expand production capacity and reduce costs, benefiting the entire downstream humanoid robot industry.
In addition, Yinshi Robot, a representative enterprise of the Shougang Fund Embodied Intelligence Ecological Base and a leading company in core components for humanoid robots, deeply participated in the energy transfer segment of the World Humanoid Robot Games, conveying innovative strength to the event. At the same time, Yinshi Robot provided complete dexterous hand hardware support and specialized technical empowerment to multiple participating teams. From dexterous hand products to technical debugging, it comprehensively safeguarded the fine manipulation capabilities of participating humanoid robots. Yinshi Robot has achieved full-process self-research and self-production of dexterous hands, and its products have been widely used in industrial, service, medical, and other scenarios. Serving a top-level event was itself a practical test of its core component localization strength.
While practical testing continued in the competition arena, industrial problem-solving advanced outside it. The dexterous hand is not only a product category. It is a platform for supply chain coordination, standards development, data collection, and model training. For humanoid robots to scale, dexterous hands must become more reliable, more affordable, and easier to integrate. They must also generate data that can improve manipulation models across many tasks. The competition, the exhibition, the factory line, and the investment fund are therefore connected parts of the same effort: moving humanoid robots from isolated demonstrations to repeatable economic value.
-
A Provincial Key Laboratory Dedicated to Dexterous Manipulation
Recently, the Beijing Science and Technology Commission and the Zhongguancun Science Park Management Committee officially released the 2026 Beijing Key Laboratory recognition results. The “Embodied Intelligent Dexterous Manipulation Beijing Key Laboratory,” led by Lingxin Qiaoshou, was successfully approved. It became the first provincial-level key laboratory in China centered on dexterous hand operation and focused on system-level innovation of end effectors. The laboratory is jointly built by Lingxin Qiaoshou, Beijing University of Science and Technology, and Beijing Information Science and Technology University. It aims to comprehensively improve the industry’s professional capability in dexterous hand operation and create an open, collaborative, globally leading innovation ecosystem for embodied dexterous hands.
In the competition arena, humanoid robots picked up a soybean with tweezers, weighed 20 grams of powder with a spoon, and inserted a USB cable with a fingertip. These seemingly small actions represent a series of unresolved bottleneck problems. The establishment of the laboratory is intended to solve these problems at the root. Embodied intelligence is a future industry laid out in China’s 15th Five-Year Plan and a key track for cultivating new quality productive forces and building new advantages in technological competition. The dexterous hand is a core板块 of embodied intelligence landing, directly determining the “hands-on ability” of robots and the depth of industrial application.
The approved key laboratory closely follows national strategy and Beijing’s industrial layout. It focuses on four core directions: high-precision perception dexterous hands, large-scale multimodal datasets, intelligent creation large models, and dexterous manipulation integrated systems and scenario applications. It concentrates on tackling bottleneck technologies such as high-degree-of-freedom precision control, multimodal perception fusion, virtual-real scenario transfer, and cross-domain skill generalization. It also aims to develop autonomous, controllable, and performance-leading dexterous manipulation equipment, systematically improving China’s professional capability in embodied intelligent end manipulation and strengthening the technical foundation for high-quality industrial development.
In other words, the achievements gained from the delicate competition events may eventually become screws tightened in factories or a cup of water handed over by a robotic hand in a home. That is the distant destination the laboratory seeks to reach. Building the laboratory is not only about breaking through a single technology. It is also about constructing a full-chain innovation ecosystem led by standards, supported by data, empowered by models, and driven by applications. The goal is to move the industry from point-by-point research to ecosystem co-construction.
On the hardware side, the laboratory relies on Lingxin Qiaoshou’s fully self-developed Linker Hand series, covering multiple degree-of-freedom technical routes. It will work with the industrial chain to formulate unified hardware interfaces, communication protocols, and evaluation standards. This is intended to end fragmented industry development and promote standardized, generalized dexterous manipulation hardware. On the data side, the laboratory will continue to expand and upgrade the DexSkill-Net dexterous manipulation dataset, pushing its volume from the million level toward the billion level. It will cover diverse scenarios such as industry, healthcare, and livelihood, providing a high-quality data foundation for global embodied intelligence research.
On the model side, the laboratory will develop an intelligent creation large model oriented toward the physical world, deeply integrating vision, touch, and action capabilities. This will strengthen the reasoning, planning, and skill generalization level of robots and empower the overall operation capability of the industry. On the application side, it will connect research and development, pilot testing, and landing, output standardized solutions, accelerate real-world deployment, and form a closed loop of technology, data, models, and applications. The ambition is to build a globally leading dexterous hand operation industrial ecosystem.
The laboratory constructs a deep industry-academia-research integration pattern featuring enterprise leadership, university empowerment, collaborative problem-solving, and shared results. The three parties complement each other’s strengths. Lingxin Qiaoshou, as a leading global enterprise in dexterous manipulation, has full-stack self-research, scaled production, and scenario-based landing capabilities. It provides the laboratory with a hardware platform, engineering support, and industrial resource guarantees. Beijing University of Science and Technology has deep accumulation in robot control and artificial intelligence basic theory, providing frontier theoretical support. Beijing Information Science and Technology University is deeply engaged in precision sensing, intelligent manufacturing, and system integration, empowering the engineering landing of technology. Together, the three parties aim to comprehensively improve the industry’s dexterous manipulation research and development level, professional capability, and cultivation of high-level innovative talent.
As the first provincial-level key laboratory in China focused on dexterous hand operation, Lingxin Qiaoshou carries the mission of breaking through core technologies, improving the industrial ecosystem, enhancing industry capability, and leading industry standards. In the future, the laboratory will adhere to the “Four Orientations,” closely align with national strategy and Beijing’s functional positioning, deepen research and development in dexterous hand operation technology and ecosystem construction, accelerate key technology breakthroughs and achievement transformation, help Beijing build a global artificial intelligence innovation highland, and contribute to the high-quality development of China’s embodied intelligence industry and a future of human-machine symbiosis.
-
Data Tables: Scale, Tasks, and Hardware for Humanoid Robots and Dexterous Hands
The following tables summarize key figures and task requirements reported in connection with the second World Humanoid Robot Games, the 2026 World Robot Conference, and the dexterous hand ecosystem. They are presented to show how humanoid robots are being measured not only by mobility but also by manipulation. The data also highlights why dexterous hands are central to the next phase of humanoid robots.
Table 1: Expansion of the World Humanoid Robot Games Metric First World Humanoid Robot Games Second World Humanoid Robot Games Events 26 51 Teams 280 666 Robots About 500 2,056 Schedule 3 days 5 days Table 2: Selected Dexterous Hand Tasks and Requirements Task Requirement Capability Tested Powder weighing Use a spoon to take and place 20 grams of powder, with error within plus or minus 0.5 grams Micro-force control, visual feedback, decision-making Bottle cap prying Use an opener to pry a cap from a glass bottle; valid openings within 5 minutes determine score Precise force in narrow space, dynamic coordination, real-time feedback Tweezers bean picking Pick up beans with tweezers Micro-force manipulation Cable connection Connect a cable Hand-eye coordination Building block assembly Build blocks within a limited time and keep the structure stable for 5 seconds Close-range vision, planning logic, grip force, precise control Electric tool assembly Assemble using electric tools Compliant force control and precise positioning Pitch-pot Pot inner diameter 13 centimeters; center point 1.5 meters from throwing line; 3 minutes standing posture Hand-eye coordination, visual recognition, trajectory planning, stable grasp, throwing force and angle, balance Table 3: Selected XHAND Dexterous Hand Specifications Model Active Degrees of Freedom Grip or Load Speed or Precision Tactile and Compatibility XHAND 1 12 Maximum single-hand grip force 80 newtons; can lift 25 kilograms Repeated opening and closing about 10 times per second; can complete more than 100 operating tasks Not specified in the reported summary XHAND 1 PRO 21 Not specified in the reported summary Repeat positioning accuracy plus or minus 0.1 millimeters; backlash plus or minus 1 millimeter 18 tactile sensors; compatible with Manus Pro and Manus Haptic Pro; supports UR and Franka robotic arm interfaces; RS-485, EtherCAT, and CAN communication protocols -
What the Competition Results Signal for the Future of Humanoid Robots
In 2026, domestic dexterous hands achieved密集 breakthroughs in indicators such as degrees of freedom, grip strength, load capacity, and tactile resolution. The second World Humanoid Robot Games listed dexterous hands as a special event for the first time, making “micro-manipulation at the fingertips” a new scale for measuring the evolution of humanoid robots. The honor of the competition belongs to the present. The industrial long march has only just begun. The real question is not whether humanoid robots can win a contest. It is whether they can reliably perform useful work when conditions change, when objects vary, and when human operators expect safe and intuitive collaboration.
As the industry develops, humanoid robots will not remain merely “able to run and jump” exhibits. They will gradually grow into machines that can extend both hands and take on real work in complex scenarios. That future depends on many layers of innovation: stronger dexterous hands, richer tactile sensing, more capable embodied models, more standardized interfaces, larger manipulation datasets, lower production costs, and a capital ecosystem willing to support long-term research and development. The dexterous hand may be small, but it carries a disproportionate share of the promise. For humanoid robots, the last centimeter of manipulation is also the first centimeter of everyday usefulness.
The path from the competition floor to the factory floor, and from the exhibition hall to the home, will be built by repeated testing, data collection, and engineering refinement. Every powder weighing attempt, every bottle cap prying motion, every pitch-pot throw, and every cable connection creates knowledge about how humanoid robots interact with the physical world. When those lessons are captured in hardware standards, datasets, and models, the dexterous hand becomes more than a component. It becomes a bridge between human intention and robotic action. For humanoid robots, that bridge is the difference between spectacle and service, between a machine that moves and a machine that works.
The second World Humanoid Robot Games and the 2026 World Robot Conference together show that the field is converging on manipulation as a decisive challenge. The dexterous hand is no longer a niche subsystem. It is a focal point for companies, research institutions, investors, and event organizers. For humanoid robots to enter real scenes, they must master the hand. They must see, feel, grasp, adjust, and finish. They must use tools as humans do, but with the reliability, repeatability, and data-driven improvement that machines can offer. That is the meaning of the last centimeter, and it is the work that will define the next stage of humanoid robots.