Sci-Tech

Embodied intelligence: from scene implementation to industrial rise

2026-03-09   

From the lively performances on the Spring Festival Gala stage to the thoughtful services in the bank lobby, from efficient operations in industrial workshops to household exercises in training centers, embodied intelligence is entering more and more production and life scenarios, promoting the deep coupling of artificial intelligence and robotics technology. This year's government work report proposes to establish a mechanism for future industry investment growth and risk sharing, and to cultivate and develop future industries such as energy, quantum technology, embodied intelligence, brain computer interfaces, and 6G. During the National People's Congress and the Chinese People's Political Consultative Conference, several representatives and industry insiders expressed that during the 15th Five Year Plan period, domestically produced humanoid robots are expected to achieve leapfrog development in various aspects such as technological breakthroughs, scene implementation, and industrialization process. They look forward to the realization of embodied intelligence as soon as possible, and the industrial ecology will become more perfect. With an increasing number of application scenarios, we look forward to being able to think and do. At Jiangsu Bank, Kuafu Robot enthusiastically explains business processes to elderly customers; In the production workshop of Zhucheng Technology in Yueqing, Zhejiang, wheeled humanoid robots skillfully transport materials and complete workshop circulation, operating in a standardized and smooth manner. From financial services to industrial manufacturing, embodied intelligence has gradually broken the development boundary between traditional robots and general models, and is seeking to achieve a leap from "thinking" to "action". Several representatives and committee members expressed their intention to explore new forms of human-machine collaboration and promote the application of embodied intelligence in positions with labor shortages and high-risk environments. Leng Xiaokun, Chairman of Leju (Shenzhen) Robot Technology Co., Ltd., believes that embodied intelligence is an emerging industry that deeply couples artificial intelligence with robots. The universal big model is the 'brain', responsible for 'thinking'; Robots are 'bodies' that solve the problem of' doing '. Embodied intelligence connects the two, enabling robots to truly achieve intelligence and universality. Humanoid robots are the best carriers of embodied intelligence. ”Zhou Jian, founder of Shenzhen Youbisxuan Technology Co., Ltd., believes that embodied intelligence can complete the entire chain of actions from perception to execution, achieving a closed loop from "being able to think" to "being able to do" compared to the general large models existing in the virtual world. However, this leap is difficult to achieve overnight. Embodied intelligence is driving humanoid robots from specialized devices that rely on pre-set programs to general intelligent agents driven by multimodal perception models. "Zhang Li, a National People's Congress representative and president of the China Academy of Electronic Information Industry Development, analyzed that in the next 3 to 5 years, humanoid robots will still need to continue learning in simulation environments and the real physical world. This process will gradually solidify the technological foundation of deep integration of" thinking and acting ". Wang Guoren, a member of the National Committee of the Chinese People's Political Consultative Conference and dean of the School of Computer Science at Beijing Institute of Technology, believes that simulation training, multimodal perception, and robot decision-making are the three core technologies that support embodied intelligence from "thinking" to "action". Among them, simulation training has become a standard technology for top enterprises, multimodal perception has been widely applied in multiple scenarios, and robot decision-making is currently in the stage of production line validation. In the view of Tang Wenbin, co-founder of the Force Spirit Machine, the value of embodied intelligence is accelerating in real scenarios: traditional industrial robots follow preset programs to "point and hit wherever", and can only stop and wait for instructions when encountering changes; And embodied intelligent robots with large modeling capabilities can autonomously judge and adjust their actions based on on-site conditions during work. The true value that embodied intelligence brings to the industry lies in the transition from 'can only execute' to 'can think and do'. Despite the vast application prospects of embodied intelligence, there are still multiple challenges on the road from laboratory to industrialization, such as the need for breakthroughs in key technologies, the need to improve the industrial chain, and insufficient data support. Several representatives and committee members suggested that it is necessary to anchor key components of humanoid robots, universal large models, and other "bottlenecks" deeply and thoroughly, while building a sound industrial collaborative ecosystem to break through industry development bottlenecks. Several experts have provided solutions to the common problem of "insufficient brain ability". Cao Peng, a member of the National Committee of the Chinese People's Political Consultative Conference and Chairman of the Technical Committee of JD Group, believes that embodied intelligence needs to connect the entire chain of perception, decision-making, and execution, and high-quality data is the key. Leng Xiaokun believes that high-quality data is not simply a record of actions, but a systematic recording of the entire process of perception, decision-making, and execution by humanoid robots when completing tasks in a real environment, which requires strict post-processing procedures to ensure data quality. The ability to collect large-scale and standardized data is an important data support for driving humanoid robots to "learn how to work". Liu Qingfeng, a National People's Congress representative and Chairman of iFlytek Co., Ltd., suggested building a humanoid robot brain using a "pre training+post training" mode of "multimodal large models+embodied intelligence" to open up a complete closed loop from perception to execution. By forming a data flywheel through real-life scenarios, it will first be implemented in high-value scenarios and gradually move towards universal applications. Leju has achieved a localization rate of over 95% for its core components by independently researching and investing in more than 10 industrial chain enterprises such as integrated joints, motors, and dexterous hands. ”In Leng Xiaokun's view, the independent and controllable hardware capabilities are a prerequisite for the industry to achieve large-scale implementation. Lai Jie, founder of Xingchen Intelligence, introduced that the company is working together with upstream and downstream to accelerate the localization and engineering of key links, promote the improvement of the supply chain and the development of industrial clusters. At the Beijing Humanoid Robot Data Training Center, the Kuafu robot is repeatedly practicing folding actions - shaking open, folding in half, and folding back in half - to optimize the industrial pattern through gradual ecological collaboration in large-scale applications. Although the actions are slow, they are already well-organized. Wang Qiang, the technical director of Leju Robot Tailored Training Center, introduced that in the future, robots can gradually take on various trivial tasks such as clothing sorting, item placement, and floor cleaning, helping people free themselves from the complexity of household chores. From factories to households, from technology validation to large-scale applications, bottlenecks are being overcome, and the dawn of large-scale implementation has already emerged. By 2026, the production capacity of Ubiquitous industrial humanoid robots is expected to reach 10000 units. ”Zhou Jian believes that at present, the supply chain of humanoid robots has taken shape, hardware costs continue to decline, and the industry is already standing at the threshold of large-scale delivery. Zhou Shuhao, General Manager of Zhucheng Technology's Home Appliance Connector Product Line, introduced that the company has introduced four Ubiquitous wheeled humanoid robots, which have been running stably in a high noise production workshop for four months. The inspection efficiency of the workshop has been improved from 3 hours per time to 2 hours. In the eyes of Qi Xiangdong, a member of the National Committee of the Chinese People's Political Consultative Conference and Chairman of Qianxin Technology Group Co., Ltd., embodied intelligence technology and market are becoming increasingly mature, and the embodied intelligence market represented by wheeled and multi legged humanoid robots is gradually opening up. Xu Xiaolan, member of the Standing Committee of the National Committee of the Chinese People's Political Consultative Conference and Chairman of the China Electronics Society, believes that embodied intelligence will promote the upgrading of AI from a "general technology" to a "scenario based application", drive breakthroughs in computing power, data, and algorithms, and form a positive cycle of "demand driven technology, technology feedback application". The increasingly mature industrial synergy ecosystem is a key support for industries to achieve independent and controllable development. ”Xu Xiaolan described the "3+3" development pattern of embodied intelligence from the perspective of industrial ecology: the top level is a leading enterprise with complete machine research and development and ecological capabilities, the middle level is a specialized and innovative enterprise, and the bottom level is a technology and basic manufacturing enterprise; Simultaneously supporting the three major systems of safety, maintenance, and standards. Several representatives and committee members expressed that from "connecting points into lines" to "gathering chains and groups", from technological breakthroughs to ecological collaboration, embodied intelligence is rapidly evolving and leaping forward, optimizing the industrial pattern, and opening up a new stage of development for human-machine collaboration and virtual real integration. (New Society)

Edit:Momo Responsible editor:Chen zhaozhao

Source:Xinhua News Agency

Special statement: if the pictures and texts reproduced or quoted on this site infringe your legitimate rights and interests, please contact this site, and this site will correct and delete them in time. For copyright issues and website cooperation, please contact through outlook new era email:lwxsd@liaowanghn.com

Recommended Reading Change it

Links