Many visitors will ask questions: What are you going to do if you both make end-user parts and launch your own humanoid robot?
This is a very frequently asked question at the WRC.
The answer to this is clear: to make a machine, to define the hand better, to turn it faster.
The end-executor, the hand of a robot, is the last centimeter of the robot's tail and the most difficult link in the hand-eye-arm-brain synchronous chain. It is difficult to fully understand all the problems that can be encountered in real work. A hand mounted on different mechanical arms, on different machines, power control, response, spatial movement, load matching all perform completely differently.
On the one hand, a large number of manufacturers, such as Intel, Intel, and Ryman, complete demonstrations and project landings by hand on their own devices, and parts business is still the main source of company revenue.
On the other hand, since the development of this type of automation platform, the essence is to build an in-house test test site. Through real tasks such as automated runway aircraft assembly, power operation, storage sorting, reverse finding the hand in the power control, life span, load, sensing shortcuts, then converting the hardware, drive algorithms.
More importantly, in many industries and special scenarios, customers need not just one hand, but a complete set of operational systems that can be delivered. For example, remote operation of power stations, auxiliary dismantling of aircraft engines, customers need a complete set of hands, arms, vision, mobile underpinnings, and control algorithms.
The logic is simple: parts serve as an industry-wide partner; the whole machine is used for grinding parts while taking on industry-customized solutions. The two are not opposites, but mutually nurturing closures.
What are the real scenes in which the famous hand can be used?
The second high-frequency question: what does the parameter look like, but what does it do outside of the demonstration?
There are many people in the industry who stay at the exhibition to show off their tea and water. But for industry customers, they are not concerned with the performance effect, but with the ability to complete the day-to-day work in production lines, warehouses and laboratories.
The well-known product matrix covers a load range of 0.1kg-150kg+ with different hand shapes with different landing scenarios, which can generally be divided into four main types of scenarios.
The first category: industrial structural scenarios, the pursuit of stability, life expectancy, cost-effectiveness; 3C-oriented assembly, disassembly of feedstock on motors, disassembly of warehouses, sorting of luggage at airports; repetition of a million or even five million cycles, completion of pick-up, transportation, on-loading of feedstock; large quantities of factories, logistics production lines for commercial use; solving the need for flexible automation in manufacturing.
The second category: high-end special manufacturing scenarios, aimed at aerospace, electrical energy. For aircraft engine leaf sorting, compact assembly dismantling, power station switching cabinets operation. These scenarios do not pursue fancy movements, but high requirements for safety, force control accuracy, fault tolerance.
Third category: Physical intelligence and scientific research scenarios, service robotics and AI algorithm research and development. Appropriate robotics to perform unstructured environmental operations, such as removing parts and cleaning up garbage.
Fourth category: Commercial and service scenarios. New retail sorting, commercial coffee production, life science laboratory automation. Facing objects of varying shape such as cups, food, experimental consumables, relying on self-adaptive packaging to clutch and complete uninterrupted collection.
As you can see, the application logic of an expert hand is not to make a hand that fits all scenarios, but to define the product according to the needs of the scene.
From factory production lines, aviation workshops, to man-made algorithm laboratories, the value of hands is not in how many human-like movements are made, but in the ability to continuously and steadily accomplish a specific task.
The third question repeatedly asked is: What is the difference between the increasing number of manufacturers in the market and the popularity of the product?
Aside from paper parameters, there are three things that WRC's customers are most concerned about: product spectrum integrity, real-time manufacturing capabilities, hardware-data-algorithm close-loop capabilities, which are key points that distinguish them from many of their peers.
Adhering to real application scenarios prioritizes, not just the pursuit of paper-based freedom. The industry has a tendency to think from a customer's point of view: how much freedom is needed, how much is done. The vast majority of industrial operations, which rely on three-finger structures to cover more than 90% of work tasks, are better than complexity and operational stability.
The country's most comprehensive end-user product spectrum, not a single bet on a single type of technical route. Many peers focus on a single five-inch sleek racetrack, whereas at the same time the know-how of the layout of flat-handed, cooperative, clever handed, heavy handed, data collector, two-, three-, five-fingered, connector, direct-drive, rope-driven technology routes parallel, load ranges from 0.1kg to more than 150kg.
After industrial-level verification, mass-production commercial capability, more than laboratory samples. Many skilled hands stay in the sample, small batch stage. The two-fingered collaborator has achieved the annual sales of tens of thousands of levels, a large number of products landing industry, human customer real projects, not just exhibitions. The product has completed ISO quality, environmental, occupational health and safety system certification, EU CE/ROHS certification, five million-level lifetime testing, from component design to Salt City manufacturing base to self-research and mass delivery.
Unique hardware + data + algorithms close the loop, not only sell hardware. Most end-user manufacturers only output hardware kernels. In addition to the handset, there is also a complete set of multi-modal data collection tool chains. Integrated and industrial project landings generate real-world physical interaction data, scale up data collection with numerical devices, data anti-mammalary drive algorithms, operating model iterations, iterations complete and update hardware products at the same time.
At the World Robotics Congress, the entire industry is looking forward to an explosion of humanoid robots, but they won't come overnight.
As many industry practitioners agree, the first step to being physically intelligent is not to be fully intelligent, but to create value.
Large models determine the upper limits of robots, while hand-held end-to-end hardware determines the lower limits of where robots land. Intelligent robots output parts outward to empower industry-wide integrated partners, while using integrated and vertical industry projects to hone hardware and data capabilities.
* Source: Knowledgeable robots