On April 19, the world's first humanoid robot half marathon came to a successful conclusion. 20 humanoid robot teams and human runners stood on the 21.0975-kilometer track in Yizhuang, Beijing.

In the end, the humanoid robot "Tiangong Ultra" crossed the line with a time of 2:40:42, the 1.2-meter-tall Songyan Power N2 Little Naughty Team won the second place, and the Walker No. 2 Robot Team won the third place.

Human-machine co-running track: an unprecedented challenge

The race adopted a "human-machine co-run" model, with human and robot runners sharing the same track although they were in different regions. For robots, this brings unprecedented challenges in terms of environmental adaptation, ground response, and communication. This is also the first time that such a format has been tried on a global scale. During this 21-kilometer journey, the bipedal robot needs to complete about 250,000 extreme tests of joint movements, involving key technologies such as dynamic balance, joint heat dissipation, and energy endurance.

Liang Liang, deputy director of the Management Committee of Beijing Economic Development Zone, said, "This competition adheres to the concept of 'participation is victory, completion is hero', because this is not only a competition, but also a test and verification platform. The 21-kilometer long distance test and the non-stop running are great challenges for many of the robot's technologies. ”

Despite the challenges, many robotics companies and scientific research institutions have actively signed up for the competition. The Tiangong Robotics team said in an interview: "The main purpose of participating in this competition is to better verify some key common technical problems of humanoid robots, and to find and solve these problems in the process." For example, the stability and heat resistance of the body and integrated joints, the generalization of motion control algorithms, and the high efficiency of energy consumption systems. "This competition is not only a survey of technology, but also an exploration of future industrialized scale application standards.

"Tiangong" robot comes from the Beijing Humanoid Robot Innovation Center (also known as the Embodied Intelligent Robot Innovation Center jointly built by the state and the local government), which is the first innovation center in China focusing on the core technology, product development and application ecological construction of embodied intelligent robots. Xiong Youjun, CEO of the center, explained the huge difference between real road conditions and experimental scenarios.

He said that in the 21-kilometer race, the robot first needs to have high-density integrated joints and machine bodies that can adapt to long-distance running and have excellent heat dissipation technology; Secondly, the robot must be able to control the coordinated movement of each joint with high precision, realize running, positioning, dynamic obstacle avoidance and other functions, and test its core algorithm development and adaptation capabilities. Finally, the "Running Horse" fully tested the stability, reliability and endurance of the robot, and put forward higher requirements for the performance of the machine in the face of complex road conditions such as turning, uphill and downhill.

|100 days will have faster iteration technology

The third runner-up in this competition, the Walker 2 robot, made history and became the first legged robot to walk more than 100 kilometers! Not only does it have a battery life of more than 6 hours, but it also uses high-strength, lightweight composite materials, which not only ensures structural strength, but also greatly reduces weight. It's like dressing the robot in a pair of lightweight but sturdy running shoes, allowing it to handle long-distance challenges with ease.

According to the leader of the Walker 2 team, the highlight of their technology is the use of tendon bionic drive technology. This technique simulates how human muscles and tendons work, but in practice there have always been problems with control accuracy. Through years of experience in motion gait algorithms, the team has successfully narrowed the gap between the practical application of tendon drive and the simulated environment, becoming the first robot to implement reinforcement learning and imitation learning based on tendon drive.

"We wanted to demonstrate the advantages of the bionic drive, in particular the stability of the tendon-rope drive in a robot marathon, the light gait and the efficient energy management," the person in charge said. Their goal is to prove the viability of this technology for long-term, long-distance operation and to promote the development of more efficient and natural humanoid robots.

The team also said that in the future, it hopes to further improve its adaptability to the dynamic environment and self-learning ability. "We wanted to explore how we could make the robot adjust its gait more efficiently and adapt to different terrains, slopes and even climatic conditions." The ultimate goal is to promote the industrialization of bionic robots. "Technological breakthroughs in the laboratory need to be industrialized in order to truly change real life. By working with industry, we can accelerate the transfer of technology and move bionic robots from the lab to the market. ”

Cui Wenhao, vice president of Songyan Power Algorithm, also expressed a similar view in an interview: "Participating in this marathon, the limit of the algorithm of whole-body motion control, the stability of the robot body, and the full-power operation of the motor have been tested under the high intensity of the marathon. ” This intensive, real-world testing is essential to validate and refine the technology.

Robotathon: the blue ocean of scientific and technological innovation

Behind this competition stand innovative teams from different fields, including companies focusing on technology research and development, as well as university research institutions. This combination of production, education and research has injected a steady stream of power into China's robot industry. Through continuous technology accumulation and iterative updates, we will see more intelligent and efficient robot products appear in various industries in the future, thereby driving the upgrading of the entire industrial chain.

Of course, achieving this vision will not be easy. At present, the robot industry is facing many challenges, such as how to improve the autonomy and adaptability of robots, and how to ensure that they operate safely and reliably in the complex and ever-changing human society. These are all issues that need to be addressed urgently. However, as this game shows, although a robot needs more than one human "runner" at this stage, it is the attitude of scientific and technological innovation to face difficulties head-on.

From precision reducers to carbon fiber joints, from brain-like chips to liquid metal sensors...... Breakthroughs in every link of the embodied intelligence industry chain are reshaping the possibilities of robots. Although the current robot marathon is more meaningful to the outside world than the running itself, when humanoid robots stand on the same starting line as humans, a new machine evolution has begun.

Jingtai investment suggestion: look at parts in the short term, look at the scene in the medium term, and bet on ecology in the long term.

Short-term (1-2 years): bet on the "robot body"

Key directions:
high-density joints (reducer, motor, carbon fiber structure) - the "knee" of the robot, who is durable wins; Sensor fusion (force, vision, touch) - the core technology that allows the robot to "not hit the wall"; Motion control algorithm – determines whether the robot is an "elegant runner" or an "iron runner".

2. Medium-term (2-5 years): focus on "scenarios that can make money"

High-potential areas:
industrial scenarios (handling, assembly, quality inspection) - clear demand and strong willingness to pay; Medical rehabilitation (exoskeletons, surgical robots) - policy support, high gross profit margin; Commercial services (hotel reception, cleaning patrols) - the logic of manpower substitution is the hardest.

3. Long-term (5 years +): Layout the "man-machine dance" ecology

Embodied Smart OS (Android's "Android"); Human-robot collaboration protocols (how to make robots and humans coexist safely).

It is recommended to track the incubation projects of head research institutes (such as Beijing Humanoid Robot Innovation Center) and pay attention to the robot layout of cross-border giants (such as Huawei and Xiaomi).