Not long ago, Logic Robotics released its first-generation general-purpose humanoid bipedal robot, "Expedition A1." This 175cm tall, 53kg humanoid robot can participate in appearance inspection processes, assemble chassis, and operate on busy automotive assembly lines. According to available information , Shanghai Logic Innovation Technology Co., Ltd. was established in February of this year, with a business scope including the research and development of intelligent robots, the development of artificial intelligence theory and algorithm software, the sales of artificial intelligence hardware , and the sales of electronic products .
Besides BYD, BlueRun Ventures, Wofu Ventures, and Qiyu Investment are also new shareholders in this round of financing for Zhiyuan Robotics. In just six months since its founding, Zhiyuan Robotics has secured four rounds of funding, indirectly demonstrating the booming humanoid robot market.
In the humanoid robot sector, companies such as Xiaomi's CyberOne, Dreame's general-purpose humanoid robot, and Cloud Ginger from CloudMinds have already emerged. Goldman Sachs predicts that the humanoid robot market may reach $154 billion by 2035, becoming another AI application scenario after intelligent driving electric vehicles .
Why are humanoid robots so popular? When did they go from being "unaffordable" to "affordable"?
What are the difficulties in industrial development?
On May 16th of this year, Tesla released a video of the Optimus humanoid robot at its 2023 shareholder meeting , showcasing its advancements in natural walking, motor torque control, force control, environmental recognition, and memory. Musk emphasized that now is an excellent opportunity to advance the development of humanoid robots because a wealth of software and hardware technologies in the field of autonomous driving can be shared . Optimus will use the same fully autonomous driving system as the company's electric vehicles. He proposed a bold vision: "Even a natural person might need two humanoid robots. In the future, the number of humanoid robots worldwide is expected to reach 10 to 20 billion."
Yao Qizhi , an academician of the Chinese Academy of Sciences and director of the Institute for Interdisciplinary Information Sciences at Tsinghua University, pointed out that the most ideal physical form for embodied general artificial intelligence (AGI) is a humanoid robot.
Humanoid robots are designed and manufactured to interact with human tools and the environment, thereby assisting or even replacing human production and daily life. Traditional robot technology often focuses on a specific aspect; for example, industrial robots focus on motion control technology, while cleaning/food delivery robots focus on navigation technology.
Compared to other service robots, humanoid robots require higher levels of perception, motion control, and interaction capabilities. Their mobility represents a system integration project, involving the fusion of various hardware and software technologies such as autonomous driving, visual navigation, sensors , servo systems, reducers, and control systems . The integration of these technologies is higher, representing an extension and evolution of existing applications.
Due to their extreme complexity and the need for massive R&D investment, humanoid robot development has been one of the most challenging aspects of robotics technology over the past few decades. As a result, pioneering humanoid robot development projects such as Honda Motor Co. , Ltd.'s "Asimo" and SoftBank Group's "Pepper" have been retired.
A research report from Minsheng Securities indicates that humanoid robots can be broken down into core components such as control systems, the robot body, and sensing systems. In areas such as voice, vision, decision-making, and control, the combination of large-scale models with humanoid robots helps the latter form a closed loop of perception, decision-making, and control, enabling the robot to possess common sense. However, the hardware required for humanoid robots, such as servos (similar to human joints), sensors, and ball screws (transmission components), still requires continued development across the entire industry chain.
A report on humanoid robots released by Huatai Securities in August of this year pointed out that the humanoid robot industry chain is still in its early stages, concentrated in upstream core components such as reducers, sensors, controllers , and chips . Downstream potential applications are vast, including logistics, security, and elderly and disability assistance. The report indicated that the key to large-scale commercialization of humanoid robots and their widespread adoption is solving the problem of high costs.
The humanoid robot industry chain consists of upstream raw materials and components, midstream robot bodies and system integration, and downstream various application scenarios. Taking servo motors , reducers, and controllers—the three core components of industrial robots—as examples, relevant research reports indicate that these core components account for over 70% of the cost of industrial robots, with reducers, servo systems, and controllers accounting for approximately 36%, 24%, and 12%, respectively. Considering the significantly increased degrees of freedom of humanoid robots compared to traditional industrial robots, the cost of reducers and motors is expected to be even higher.
Why is it so popular?
Several companies have recently launched new humanoid robot products, and at the same time, an increasing number of companies incubated by university research institutions are moving humanoid robots from the laboratory to commercial applications. One of the reasons behind this is the development of large-scale artificial intelligence modeling technology, which provides technical support for the industrialization of humanoid robots. Industry experts, including NVIDIA CEO Jensen Huang, believe that the next wave of AI will be embodied intelligence.
Currently, the hardware itself has reached a bottleneck due to physical limitations imposed by its electric drive system. The real competition lies in whether robots are smarter, that is, the artificial intelligence capabilities of large-scale models. Robin Li, founder, chairman, and CEO of Baidu, once stated that large-scale models have successfully compressed human cognition of the entire world, showing us the path to achieving general artificial intelligence. "Currently, we are at a new starting point, a new era of artificial intelligence centered on large-scale models. Large-scale models have changed artificial intelligence, and large-scale models are about to change the world."
In June of this year, the Ministry of Science and Technology released the "Research Report on China's Large-Scale Artificial Intelligence Model Map". The report showed that, in terms of the distribution of large-scale models released globally, China and the United States are far ahead, accounting for more than 80% of the global total. The United States has always ranked first in the world in terms of the number of large-scale models.
How can general-purpose robots achieve low-cost mass production and generate practical application value? Peng Zhihui, co-founder of Zhiyuan Robotics, known as "Zhihui Jun," stated that hardware is only one prerequisite; the more important element is the robot's underlying brain. Zhiyuan's EI-Brain architecture divides the robot's thinking system into several layers: a cloud-based superbrain, an edge-based brain, a cerebellum, and a brainstem, corresponding to the robot's task-level, skill-level, instruction-level, and servo-level capabilities, respectively.
Yao Qizhi stated that the emergence of large-scale artificial intelligence models will create more new value for industries, enabling the development of a general-purpose robot with broad application capabilities. This requires the robot to possess the three key components of a human embodied intelligent agent: a brain, a cerebellum, and a body, while also being better adapted to human social environments. For embodied intelligent robots, the "brain" governs upper-level logical reasoning, decision-making, planning, and communication with other intelligent agents and the environment using natural language; the "cerebellum" primarily controls the body by dominating various senses such as vision and touch to complete complex tasks; and the "body" requires sufficient hardware, such as sensors and actuators.
Many companies have already begun to launch dedicated large-scale models for the robotics field. For example, Google's DeepMind launched the RT-2 model to improve robots' natural language understanding capabilities; and domestic player CloudMind released RobotGPT, the industry's first multimodal large-scale robot model. This means that the application of general artificial intelligence in humanoid robots will enhance capabilities in simulation and data generation, task decomposition and planning, and autonomous operation and control, demonstrating enormous application potential.
Dongwu Securities stated that with Tesla's release of its humanoid robot product last year, the commercialization of humanoid robots has accelerated. Leveraging Tesla's leading role, global commercialization of humanoid robots is expected to begin in 2025, with sales reaching 20,000-30,000 units. By 2030, additional demand is projected at 1 million units, and sales are expected to exceed 10 million units by 2035.
Approaching the tipping point of transformation and leapfrog development
Currently, the robotics industry is gradually approaching a critical point of transformation and leapfrogging. The "China Robotics Technology and Industry Development Report (2023)" points out that common and cutting-edge technologies in robotics are developing rapidly, and a large number of sciences and technologies such as electronics , mechanics , biology, and materials are mutually reinforcing each other through integration. Integrated innovation, high-end development, intelligence, and ecosystem construction have become the key words for the future development of robotics in China.
From the perspective of the humanoid robot industry, its three core technologies mainly include human-computer interaction, scene perception, and motion control. Among them, scene perception technology has made relatively rapid progress. In terms of motion control, most of them adopt hydraulic drive or pure motor drive. There is still a lot of room for improvement in motion control algorithms. In terms of human-computer interaction, there have been breakthroughs with the development of artificial intelligence large models, but there is still a big gap from "autonomous decision-making".
According to Yao Qizhi, embodied intelligent robots currently face four challenges: first, they cannot achieve a one-step implementation with a large embodied model like a large language model; second, there are challenges from computing power; third, how to integrate multimodal senses; and fourth, the need for data collection.
Currently, humanoid robots are limited in application scenarios due to technological and cost factors, but they possess boundless potential. Policy is strongly supporting this. On August 22, the Ministry of Industry and Information Technology and three other departments issued the "Implementation Plan for the New Industry Standardization Leading Project (2023-2035)," focusing on standardization work in emerging and future industries, forming "8+9" key areas for new industry standardization. Among them, the future industry focuses on nine major areas, including humanoid robots and generative artificial intelligence. Recently, the National Robotics Standardization Technical Committee announced the establishment of the Humanoid Robot Standardization Working Group and appointed iFlytek, Xiaomi Group, and Zhejiang Lab as deputy leaders of the working group to jointly promote the standardization of humanoid robots in China. At the local level, the "Several Measures of Beijing Municipality to Promote the Innovative Development of the Robotics Industry" proposes to organize and implement a "foundation-building" project for the robotics industry and release a list of key technologies for industrial breakthroughs. It will also establish a 10 billion yuan robotics industry fund, with an initial scale of no less than 2 billion yuan.
At the 2023 World Robot Conference held recently, Xu Xiaolan, Vice Minister of the Ministry of Industry and Information Technology, stated that new technologies, new products, and new business models, represented by humanoid robots and general artificial intelligence, are booming and becoming the commanding heights of global technological innovation, new tracks for future industries, and new engines for economic growth. The strong momentum of China's robot industry has laid a solid foundation for the development of the humanoid robot industry.
Currently, the application fields of robots are expanding rapidly, and their application in areas such as new energy vehicles, medical technology, logistics warehousing, outdoor delivery, and commercial services is continuously deepening. From a technological perspective, cutting-edge technologies such as artificial intelligence, biotechnology, new materials, and brain-computer interfaces are developing rapidly. As carriers and platforms of these cutting-edge technologies, robots are accelerating their deep integration with these technologies, bringing profound impacts on the form, function, and application scenarios of robots.
BYD's investment in Zhiyuan Robotics demonstrates that BYD can utilize humanoid robots for tasks such as appearance inspection and chassis assembly. On a deeper level, this industry chain layout may drive the research and application of technologies like autonomous driving. Regardless of BYD's future plans for humanoid robots, it's certain that the humanoid robot industry will experience rapid development.
Regarding whether humanoid robots will eventually enter every household, Zhou Jian, founder, chairman, and CEO of UBTECH Robotics, believes that it's only a matter of time before they become part of home companionship services. However, he also stated that humanoid robot task planning involves technologies, common sense, and knowledge from multiple fields, and combining these capabilities to accomplish human tasks will be a significant challenge.
Whether it's automakers like Tesla, Honda, and Toyota, or giants like Huawei, Xiaomi, Samsung , Google, and Amazon , achieving cost reduction and supply chain layout remains a crucial hurdle that must be overcome in the future development of the humanoid intelligent robot industry.
