Behind the championship win of the "shy" robot, Xiong Youjun talks about the next stage of development for humanoid robots.
On August 26, the 2nd World Humanoid Robot Games concluded at the "Ice Ribbon" in Beijing.
Compared with the first edition, humanoid robots run faster, perform more complex actions, and can complete an increasing number of tasks. At this Games, the results of multiple events including 100-meter, 400-meter and 150-meter races have been greatly refreshed.
Many robots have thus gained attention beyond the industry circle.
On the athletics field, the Tiangong Ultra from Beijing Humanoid Robot Innovation Center won the championship of the 400-meter large group with a time of 38.15 seconds, and then ran 8.64 seconds in the 100-meter final; the Tiangong Omni, which once became a hit online for its "shy" movements, won the championship of the 400-meter small group with 45.66 seconds. The Zhuifeng Zaizai team of Honor's "Lightning" robot won the second place in the 400-meter large group, with a result also within 40 seconds.
In more complex events, the tests faced by robots are also increasing. The Zhiyuan Lingxi X2 won the championship of the 100-meter obstacle race, which requires continuously completing obstacle avoidance, crossing and narrow-space passage; the accelerated evolved Booster T2 has entered multi-person confrontation events such as football, where robots need to independently complete positioning, collaboration and real-time decision-making.
The competition field also exposed some unsolved problems. After running at high speed, some robots still find it difficult to decelerate quickly, and even directly hit the protective pad; in multi-person confrontation events such as football, there is an obvious gap between the robot's collaboration, judgment and on-site response capabilities and the performance required for a truly smooth competition.
To a certain extent, this Games is like a centralized physical examination of the capabilities of humanoid robots. On the one hand, capabilities such as speed, balance and whole-body coordination are being rapidly improved; on the other hand, environmental adaptability, autonomous decision-making and stability in complex tasks are still being tested.
However, the competition field is only a capability test. After entering real scenarios such as factories, logistics and energy, robots are faced with more stringent tests on reliability, efficiency and cost.
In the past few years, more and more robots have begun to enter scenarios such as automobiles, 3C, logistics and retail. The industry's evaluation criteria for humanoid robots are also constantly changing, and the evaluation dimensions are gradually extending from single technical performance to stability, efficiency and commercial return.
This also makes the competition among robot companies extend from individual capabilities to the whole system. Ontology and motion control are still the foundation, and the importance of large models, operation capabilities, engineering, mass production, delivery and after-sales services is rising. After the robot truly enters the production environment, any short board in any link may affect the final implementation effect.
The Tiangong Ultra, which won the championship of the 400-meter large group at the Games, and the Tiangong Omni, the "shy" champion of the 400-meter small group, both come from the Beijing Humanoid Robot Innovation Center. Recently, Xiong Youjun, General Manager of the center, talked about this ongoing change in the robot industry when communicating with the media.
In his view, in the past, the industry paid more attention to technical parameters, but now it is paying more and more attention to a company's "systematic capability": from R&D, to factory production, to customer experience and services, the requirements for corporate technology and organizational capabilities are all increasing.
While the performance of robots on the field is still being continuously refreshed, the industry has begun to ask the next level of questions: how these capabilities can be transformed into stable, replicable and economically viable productivity after leaving the competition field.
The following is the selected record of the communication between Xiong Youjun and the media:
01
When Humanoid Robots Start Calculating ROI
Q: In the past, when the industry evaluated humanoid robots, it often paid attention to individual parameters such as speed, load and degree of freedom. Tiangong has won multiple championships at the Humanoid Robot Games this time. As robots begin to enter real scenarios, what changes do you think have taken place in the industry's criteria for evaluating a robot company?
Xiong Youjun: I think a relatively obvious change is that everyone may have paid more attention to technical parameters in the past, and now they are paying more and more attention to the systematic capability of a company.
This includes not only R&D capabilities, but also the ability to truly turn technology into products, form factory production and large-scale delivery, as well as subsequent customer experience, service capabilities and the company's sustainable development capabilities.
Therefore, for a robot company, the current challenges are more comprehensive, and the requirements for technical comprehensiveness and the overall organizational capability are all increasing.
Q: Now there is a skeptical voice at home and abroad, believing that humanoid robots seem to be very popular, but a large number of demonstrations are still concentrated in scenarios such as singing, dancing and competitions. How do you view the relationship between these demonstrations and the real production value created by robots?
Xiong Youjun: I think truly professional people will not make excessive judgments on robots just because they can sing and dance.
Robots ultimately need to serve human beings. One of its important values is to enter scenarios that people are unwilling to do or that are harmful to people, and assist humans to complete these tasks.
Of course, singing and dancing itself can also provide certain emotional value, which is no problem.
The current robot industry is still in a relatively early stage, and some companies' product positioning may be more biased towards stage display or performance. These products can also accumulate data and iterate products in the process of display and use, laying a foundation for subsequent product capability improvement, which I think is normal.
But in the long run, we still hope that robots can be more targeted at the scenarios that humans are in urgent need of, and solve practical problems through continuous product iteration.
Q: Now more and more robot companies are able to enter factories to complete POC, but there are still limited projects that have achieved continuous deployment and large-scale replication. From the scenarios you have entered, what is the most critical change for robots to move from technical verification to industrialization?
Xiong Youjun: I think it is only the very beginning that robots truly enter factory practice.
A very obvious change is that customers have begun to truly pay attention to ROI, that is, return on investment.
I think this is actually a very important sign of the gradual maturity of the industry. When everyone starts to calculate input and output, it will in turn promote robot enterprises to continuously reduce costs, improve reliability, and improve the efficiency of products and serving customers.
This requires the whole industry to jointly accelerate iteration.
Q: If we compare humanoid robots with the already very mature industrial robots and automation equipment, what is the gap in ROI at this stage? Which capabilities are mainly needed to narrow this gap?
Xiong Youjun: At present, if we calculate ROI, there is indeed a certain gap between humanoid robots and traditional industrial robots.
But what I see is that this gap is narrowing very rapidly.
This will also prompt relevant enterprises to continuously reduce costs, improve the reliability of robots, and improve the efficiency of serving customers. I think all these require the joint promotion of the whole industry to accelerate the iteration of products and technologies.
Q: From Tiangong 1.0, 2.0 to the current Tiangong 3.0 and Omni, you have always maintained a relatively fast ontology iteration speed. At this stage, what changes have taken place in the product definition of the Tiangong platform?
Xiong Youjun: The Innovation Center currently has two major platforms, one is the Tiangong hardware platform, and the other is the Huisai Kaiwu general embodied intelligence platform. Both platforms are essentially open to the industry.
Tiangong is more of a general-purpose robot platform. The Tiangong Omni released this time hopes to provide a more suitable platform for industry operation and vertical applications, while further considering cost, mass production and industrialization.
Therefore, from the perspective of the ontology platform, we will pay more and more attention to the problems that need to be solved after robots truly enter the industry.
In the future, we will also invest more resources in Huisai Kaiwu, which is the "brain" and "cerebellum" platform of robots, hoping to provide more general capabilities for the industry at the level of robot intelligence and control through continuous iteration.
02
The Next Stop of Humanoid Robot Competition
Q: When more and more ontology, model and component companies enter this industry, standards are becoming the infrastructure for industrial scaling. What are the most urgent standard issues that need to be solved for humanoid robots at this stage?
Xiong Youjun: At present, the Ministry of Industry and Information Technology has established a relevant standardization committee, and is also stepping up the establishment of different professional working groups to promote the formulation of standards.
The Innovation Center also takes the lead in participating in some standard work. Now we have participated in the formulation of dozens of relevant standards, including some international standards that are being promoted.
From the perspective of the Innovation Center, we are currently focusing on several directions.
The first is safety standards. For robots to truly enter factories, commercial scenarios and even families in the future, safety is a very important foundation.
The second is interface standards. Through interface standards, the versatility between different systems can be improved, and repeated development by enterprises can also be reduced.
In addition, other working groups are promoting standards in different directions. On the whole, the current speed of industry standard construction is relatively fast.
Q: China has now formed a relatively complete robot supply chain and a large application market. Is it possible to further transform the industrial scale and scenario advantages into the right to speak in international standards in the next stage?
Xiong Youjun: This is a very important thing.
With the continuous development of China's robot industry, these standards will help Chinese enterprises better enter the international market on the one hand, and enhance the influence of China's robot industry in the international standard system on the other hand. We are now promoting some standards to the international direction.
As the capability of China's robot industry continues to grow, our participation and influence in international standard formulation will be further improved in the future.
Q: Recently, robot enterprises have received high attention in the capital market, and relatively representative cases have emerged, such as the listing of Unitree Robotics. How do you view the impact of capitalization on the whole industry? As a national-level innovation platform, how will the Innovation Center consider its own sustainable development in the future?
Xiong Youjun: From the perspective of the whole industrial development, this is a good thing. Everyone will see the future development space of this industry, and more resources may enter the robot industry.
The Innovation Center is essentially a company, and it is also a national-level innovation platform that undertakes some national missions. So first of all, we must fulfill the national mission, and at the same time hope that the company can achieve sustainable development.
As an embodied intelligence innovation platform jointly built by the state and local governments, what we need to do first in the early stage of industry development is core technology breakthrough and technological innovation.
Second, we will further open up these technologies, so that more enterprises and upstream and downstream companies can continue to innovate on our basis, and try to avoid reinventing the wheel.
Through core technology supply and platform empowerment, more Chinese robot enterprises can participate in international competition faster and occupy the technical high ground. I think this is what the national innovation center should do.
Q: From your observation, what substantive changes have taken place in the position of Chinese robots in the global industry in the past two years?
Xiong Youjun: I don't have the specific export data of robots, which may need to refer to the data of the customs. But from my own observation, there are several obvious changes.
The first is in the academic field. Now when we participate in some international conferences, we can clearly feel that more and more Chinese engineers, scientific and technological workers and university teachers are giving presentations at the conferences.
The technologies and platforms studied by many Chinese teams are already at a relatively leading level in terms of depth and height in the industry. Behind this, a relatively strong industrial foundation is actually needed for support.
The second is that China now also has strong international competitiveness in cutting-edge technology research. For example, in terms of robot-related patents, the number of applications and the proportion of China are already very high.
In addition, from the perspective of industrial scale, the supply capacity of China's robot industry is also getting stronger and stronger. Therefore, no matter in terms of scientific research, technology or industrial capacity, the influence of Chinese robots in the world is continuously improving.
This article is from the WeChat official account "Tencent Tech", written by Li Hailun, edited by Xu Qingyang, and published with authorization from 36Kr.