Industrial humanoid robot is a new term that has been very popular recently, simply put, it is a humanoid robot that operates in industrial scenarios.
The humanoid robot industry has entered the commercial testing stage from the early technology verification stage, and the commercial potential brought by it is worthy of the attention of enterprises.
▍The landing needs of industrial scenarios are clear
Recently, Toyota has tried to cooperate with Boston Dynamics to bring humanoid robots based on large AI models to the assembly line of automobile factories, and Boston Dynamics plans to deploy task-focused humanoid robots in Hyundai car factories "in the next few years", and at the same time cooperate with Toyota to research the long-term goal of humanoid robots based on artificial intelligence multitasking.
Coincidentally, UBTECH recently officially released a new generation of industrial humanoid robot Walker S1, which has entered BYD and other automobile factories for practical training, and works in collaboration with L4 unmanned logistics vehicles, unmanned forklifts, industrial mobile robots and intelligent manufacturing management systems to help realize the unmanned and large-scale commercial landing of indoor and outdoor logistics scenarios. This is also the world's first industrial scenario solution for humanoid robots and unmanned logistics vehicles.
Tesla's humanoid robots are also entering the car factory, and there are already 2 humanoid robots Optimus working at Tesla's Fremont factory, which can also work autonomously, they can remove the battery cells from the end of the production line and put the battery cells in the transport box, and more Optimus is expected to work in Tesla factories next year.
Automakers have been at the forefront of automating more assembly line processes to cut labor costs and improve worker safety. According to the latest data from the Frankfurt-based International Federation of Robotics (IFR), the automotive industry is second only to electronics manufacturers in the number of robot assistants deployed in factories globally, with about a quarter of new robotic devices added to the automotive industry in 2022.
According to the National Bureau of Statistics, as of June 2024, the cumulative number of workers employed in China's auto manufacturing industry was 4.54 million. The total number of employees announced in BYD's 2023 annual report is 703,504, of which 524,673 are production personnel, accounting for about 74.6%. According to this estimate, there are about 3.39 million production personnel in China's automobile manufacturing industry alone, and assuming that the replacement rate of humanoid robots replacing workers is 10%, then China's automobile manufacturing industry has a potential demand for about 340,000 humanoid robots.
In foreign automobile factories, industrial humanoid robots have been recognized as a competitive solution. Humanoid robots have a strong substitution for labor, overseas labor costs are expensive, and the labor union is strong, if the labor union power is considered, the current average salary of workers in United States automobile factories is 24 US dollars/h, and the weekly salary of full-time employees is 1100 US dollars (≈ 27.5 US dollars/h).
In some well-known car factories, these workers earn higher wages, such as $34.8/h for a Toyota United States factory worker, and Germany Volkswagen workers start at €13.5/h ($14.7/h ≈).
The hourly wage of the robot is much lower than that of overseas labor, and it can theoretically work 24 hours: assuming that the robot depreciates for 5 years, works 350 days * 24 hours for 1 year, and annual maintenance and repair is 15% of the price of the machine, if the price of the humanoid robot is 100,000 US dollars, the cost is 4.2 US dollars/h, and if the price is 20,000 US dollars, the cost is 0.83 US dollars/h.
For the factory, this is 2~3 years to be able to recover the cost, and the comprehensive ROI is already worth considering.
The demand for industrial humanoid robots is also very clear in China, and taking into account efficiency and labor costs is a continuous concern for enterprises. According to the "Manufacturing Talent Development Planning Guide" released by the Ministry of Human Resources and Social Security, the Ministry of Industry and Information Technology, and the Ministry of Education, by 2025, the shortage of manufacturing workers in China will be close to 30 million, with a gap rate of 48%. Among them, the three fields of handling and loading and unloading, welding and brazing, assembly and disassembly have always been the areas with large gaps.
Many manufacturers see that humanoid robots, with their advantages of being more adaptable to unstructured environments, are expected to alleviate the demand for labor in the manufacturing industry in the future. However, at present, the opponent efficiency of industrial humanoid robots is still very high, and these workers can basically complete a series of complex tasks such as picking up parts, finding the right position to complete assembly, and reviewing quality within 50 seconds.
At present, there are not many places where humanoid robots are stronger than humans, especially in terms of efficiency. But it is loyal, more serious than humans, does not desert, and is constantly learning and improving, does not go on strike, and is more flexible.
▍How to make humanoid products in the factory
At present, the industrial humanoid robots of global manufacturers are still in the pilot stage. According to general industry estimates, an annual output of 100,000 units is a small batch production, 1 million units is to enter the mass production period, and hundreds of units are just industrial pilots.
Then, in the case of pilot to small batch production, the core determining factor of entering the scenario is not the price of the robot, but the performance of the robot, the continuity and stability of production, etc.
For example, UBTECH is very clear that the core competitiveness of industrial humanoid machines lies in the accumulation of actual combat data and operational capabilities, and based on a clear commercial cognition, UBTECH is also one of the few enterprises that has opened up the whole chain of technology-production-terminal scenarios.
On the one hand, UBTECH began to take the lead in cooperating with a number of car companies to explore industrial applications, taking the lead in the world with Dongfeng Liuzhou, Geely Automobile, FAW Hongqi and other automobile companies, and has cooperated with many industry leaders including Foxconn and SF Express across industries, accumulating sufficient real data in actual combat, building a large model containing multiple types of character scenes to fine-tune its own data set, using the relatively standardized architecture scheme of large model + small model to optimize the reliability and cost control experience of robots, and can effectively iterate according to customer needs in a timely manner.
On the other hand, UBTECH has designed the third generation of dexterous hands with pressure monitoring function, which optimizes the movement performance of legs and feet, allows humanoid robots to more flexibly cut into relatively high-end jobs such as fingertips, and at the same time opens up the factory MES to improve the ability to compete, and its application has a certain typicality.
In the field of logistics, as industrial manufacturing enters the 4.0 era, more and more industrial enterprises are focusing on flexible manufacturing, and the demand for customized solutions in the field as a whole is higher, which will also promote warehousing and logistics to be fully intelligent and flexible. Moreover, warehousing and logistics is also an industrial scenario with a relatively high automation rate, and there are a lot of tedious and repetitive tasks in the logistics industry, such as handling materials and boxes.
In the past few years, Digit has upgraded from V1 to V4, which has also strengthened its job execution capabilities, not only replacing its hands with end-effectors that can be used to move containers, but also optimizing its head and eyes to support human-machine tracking and interaction, and at the same time opening up the MES system.
This approach is no different from the logic of the original warehousing and logistics robot cutting into the market.
Some time ago, Zhiyuan Robotics signed a strategic cooperation agreement with PIA, which also plans to realize data accumulation, algorithm iteration, scene generalization, etc. in different intelligent manufacturing scenarios on the basis of the "heterogeneous intelligent industrial Internet of Things" built by PIA, and realize the collaborative production guided by big data, implement refined management of the production process through digital twins, and complete the continuous iterative upgrading of robot software and hardware systems. Zhiyuan Robotics plans to complete the mass production of humanoid robots in October.
The K2 humanoid robot released by Kepler today is also aimed at industrial logistics scenarios, entering industrial warehouses to practice handling bins, sorting and express delivery, etc. The difference is that it can realize embodied operation by adding remote operation, and the actual dexterous hand performance of the robot is superior, not only able to lift the box with both hands, but also able to grab the SF Express bag, which brings a good market imagination.
Industrial humanoid robots are different from the home version of humanoids, and their positioning is mainly built for industrial scenarios, which needs to have typical industrial attributes.
Industrial scenarios are generally cost-oriented, and if the control effect and cost-effectiveness of humanoid robots are not as good as those of human labor after landing in industrial scenarios, it will affect the application willingness of industrial scenarios, or limit the data collection and iteration speed of robots, and adversely affect the development of general-purpose humanoid robots.
Therefore, at present, most of these companies have learned to start from the industrial scene and look for the commercialization possibility of humanoid robots, and the continuous strengthening of the robot's operational capabilities has always been the core.
At this stage, the humanoid robot entering the factory can summarize several typical development directions: safety, operational reliability, long-term stability, practicality, high load, flexibility. Manufacturers try to find a balance between profit and selling point between one or more of them.
▍ Conclusion and the future
Some institutions expect that if the global small batch production reaches 100,000 robots, it is expected to replace 1/10 of the long-tail working station, and under the output of 1 million units, robots can not only meet the needs of intelligent manufacturing factories, but also begin to enter the service industry and even families.
At this time, the reduction of hardware prices and the generalization ability of software will truly become the core competitiveness of enterprises in the new stage.
Because with the increase of support on the policy side and the support of large models on the software side, it is expected that with the iteration of Al technology, the ability of the robot software layer will be upgraded, and the hardware side will be rapidly iterated on, so that the accuracy requirements of the core hardware will be reduced, and the generalization of the scene will be strengthened, thereby driving the local cost reduction, and the supply side will emerge more amazing products, and the overall cost will be reduced to lower than the cost of a single human being.
Humanoid robot manufacturers need to get rid of the thinking of "selling shovels", and can really go to the scene to find industrial gold mines.
This has been pointed out in the interpretation of the Ministry of Industry and Information Technology at the press conference held by the State Council Information Office: deepen the application of scenarios, cultivate and expand new quality productivity such as low-altitude economy and intelligent manufacturing, vigorously develop new fields and new tracks such as humanoid robots, brain-computer interfaces, and 6G, support enterprises to create an "upgraded version" of intelligent manufacturing, focus on the application of intelligent manufacturing technology in typical industries, and gradient cultivate a number of basic, advanced, excellent, and pilot-level smart factories.
Creating value is the last word in the current humanoid robot products.