Beijing, Dec. 13 (Xinhua (Reporter Peng Qian Zhang Manzi) Following the large model, "embodied intelligence" has become a new hot spot in the science and technology industry this year, and is considered to be the key direction in the new wave of artificial intelligence (AI). Many startups have emerged, financing has reached new highs, and technology has continued to break through... Humanoid robots, as the most representative entity in the field, are accelerating industrialization under the catalysis of large models.
If the large model is regarded as an "interesting soul", the humanoid robot enabled by "embodied intelligence" has a "good-looking skin" and has become a powerful assistant to humans in many fields.
A humanoid robot "Qijiang 2" grabs an orange and hands it to a visitor at the exhibition hall of the 2024 World Manufacturing Congress on Sept 20, 2024. (Photo by Huang Bohan/Xinhua) Redefining robots
As a bridge between virtual and real space, "embodied intelligence" refers to the integration of AI into physical entities such as robots, giving them the ability to perceive, learn, and dynamically interact with their environment like humans.
The term "embodied intelligence" itself has a strong technical philosophy color. In 1945, French philosopher Maurice Merleau-Ponty proposed the concept of "embodied", which believes that humans need to interact and perceive the surrounding environment through the body to understand the world. In 1950, the British computer scientist Alan Turing, known as the "father of AI", first proposed the concept of "embodied intelligence" in his paper "Computing Machines and Intelligence".
In fact, industrial robots (robotic arms) with a relatively low level of intelligence have long been widely used in the manufacturing industry, bringing quality and efficiency improvements. However, the traditional industrial robot is a combination of "fixed program + robotic arm", while the robot enabled by "embodied intelligence" is an iteration of "multi-modal perception + brain decision-making".
Xu Huazhe, assistant professor of Cross Information Research Institute of Tsinghua University, believes that the future robot will present a variety of forms: two-legged, four-legged, wheeled, mechanical dogs, intelligent drones and even mechanical bees, but humanoid robots have the best adaptation to human society and will become the most able to help human robots.
Humanoid robots can solve the "last mile" problem of the production line. Many personalized and customized products cannot be unified by assembly line, which requires humanoid robots with generalization ability to "help", and assemble mass-produced parts into products according to customers' customized needs. In more complex and changeable scenes such as family services and public services, humanoid robots are also more advantageous and can adapt to different environments and needs to complete a variety of tasks.
Three difficulties to be overcome
The research and development of humanoid robots begins with the learning and imitation of human beings, and its difficulties can also be understood by comparing with the human brain, cerebellum and ontology. The "brain" is the center of the robot responsible for autonomous learning, planning and decision-making; The cerebellum is responsible for motor control, from walking to running and jumping, and from simple grasping to complex hand movements. The "body" part includes the trunk and limb structure and the dexterous hand design.
Jiao Jichao, vice president of Youbei Science and Technology and president of the Research Institute, told reporters that at present, there are more technical difficulties to be broken through in these three areas: "brain", cloud edge integrated computing architecture, multi-modal perception and environmental modeling are the focus of technology in recent years, "the biggest difficulty in human imitation is the imitation of the human brain, the existing scientific theory of the human brain research is far from enough"; In cerebellum, human-computer interaction ability, complex terrain passing, whole-body coordination and fine work are important directions. In terms of "ontology", key technologies such as rigid-flexible coupling bionic transmission mechanism, high-compact robot limb structure and dexterous hand design are important hardware bases for humanoid robots to move flexibly.
The emergence of large models has made the robot "brain" significantly "evolve", greatly improving the versatility and generalization of robots, and is expected to reduce the development cost of humanoid robots and accelerate its entry into thousands of households.
According to Yang Fengyu, founder and CEO of Ulich Robot Technology Co., LTD., the industry now mainly uses pre-training large models to pre-train robots, so that they have stronger learning ability; The large model can transfer the task-specific learning to the robot task and improve its adaptability. In addition, the multi-modal processing capability of large models can be used to combine various inputs such as vision, hearing, and touch to improve the understanding of complex scenes.
Photo taken on Aug 21, 2024 shows a bionic humanoid robot at the World Robot Conference 2024. [Photo by Ren Chao/Xinhua] China does not lag behind
Looking around the world, humanoid robots have entered the initial stage of industrialization, and have begun to "test the water" in the fields of industrial manufacturing, commercial services and family companionship. Whether it is technological breakthrough, landing progress or financing scale, the humanoid robot research and development competition is basically dominated by China and the United States.
Walker humanoid robot this year focused on the automotive, consumer electronics and other key areas of manufacturing, has entered a number of car factory training; Yushu Technology's robot has achieved a completely human-like natural walk; Ulich robot is brewing "into the home" plan; Boston Dynamics' new Atlas robot enables flexible moving of parts between different lockers in the factory; Tesla's humanoid robot "Optimus Prime" plans to start mass production in 2025...
"If the humanoid robot industry is compared to a marathon, China, Europe and the United States are almost at the starting stage of the first 1,000 meters," Jiao said.
Yang Fengyu agrees that the ability to develop large models and lead in perception technology makes US companies more competitive in robotic decision-making systems and complex task processing. China's advantages are more reflected in the field of industrial robots, especially in the manufacturing industry, the application is more mature. China has also taken the lead in patents for humanoid robots.
Although there are still challenges in the core algorithms and high-end chips of the robot "brain", the Chinese robot industry has a rich application scenario and a large potential user population, and data is one of the biggest competitive advantages.
How to collect large amounts of data in real world scenarios, and coordinate the data circulation and supply to the humanoid robot industry is the key. Xu Huazhe introduced that the real world data is generally collected by the robot companies themselves. The Beijing Humanoid Robot Innovation Center is planning to create an open source data set for use by academia and industry. In the future, high-quality shared data sets will greatly help the industry development.
