Easy material handling in factories, pre diagnosis for patients in hospitals, and racing on the field... From industrial production to livelihood services, embodied intelligence is accelerating its integration into various industries. Experts believe that by 2025, the embodied intelligence industry ecosystem will continue to improve and policy support will continue to increase. The next 3 to 5 years will usher in a critical period of large-scale development.

Diversified industry applications

　Recently, the first batch of hundreds of full-size industrial humanoid robots Walker S2 from Ubiquitous have started mass production and delivery, and will be put into frontline industrial applications in batches. In the video of the mass production delivery site released by Ubiquitous, the robots lining up for inspection move in unison, bringing a shock that only science fiction films can have. This delivery is not only a crucial step for Ubiquitous to achieve its goal of delivering 500 units within the year, but also marks a new stage in the large-scale application of its humanoid robots.

　　I can't do anything well, "a 6-year-old girl said to the robot" Dudu ". In the past, artificial intelligence may have comforted her with the phrase 'Come on, you're great!'. But this time, the "Doo Doo" equipped with JD JoyInside 2.0 chose another response: "Do you still remember the 'rainbow tail puppy' you drew?" At that moment, the frustration in the girl's eyes gradually dissipated and she began to excitedly describe her artwork. This is not a preset dialogue, but a true reflection of JoyInside's "emotional continuity" and "memory".

　　Since the beginning of this year, the speed of cultivating embodied intelligence has accelerated. Pu Songtao, director of the Future Industry Research Center of CCID Think Tank, analyzes that the product forms are more diversified, robots in various forms such as humanoid, wheeled and quadruped are constantly emerging, and product innovation in fields such as unmanned aerial vehicles, unmanned boats and autonomous vehicle continues to be active. Industry applications are also becoming more diversified, and the exploration of application scenarios of embodied intelligence in industrial manufacturing, home services, medical rehabilitation, low altitude economy and other fields is gradually accelerating.

Zhao Gang, the director of the Saizhi Industry Research Institute, believes that the level of embodied intelligence has significantly improved. Compared with traditional robot applications, intelligent robots embed large model "brains", which significantly improve their intelligence level in multimodal information perception, complex scene reasoning, and real environment action interaction. At the same time, the purchase cost has significantly decreased. Benefiting from moderate expansion of production capacity, independent and controllable key components, and standardized products, the cost of embodied intelligent products has significantly decreased. Under the influence of favorable factors, investment in the embodied intelligence industry will be more active in 2025, product technology will accelerate iteration, and the market size will grow rapidly.

Huge development potential

　Proverse Robotics recently released the industrial grade wheeled embodied robot Proverse White 2.0 version. The first generation Proverse Big White Robot, launched in August this year, has completed more than 10 landing verifications in three types of scenarios with its dual advantages of "flexibility+accuracy". The second generation product has achieved comprehensive upgrades in more than 300 performance points. Founder and Chief Operating Officer of Proverse, Ge Jin, believes that embodied intelligence has penetrated into multiple aspects of industrial manufacturing, but its value is far from reaching its peak. To truly unleash its potential, it is necessary to overcome core processes that require extremely high precision, quality, and efficiency.

　　The embodied intelligent products mainly solve the pain points of traditional robot base fixation, inflexible operation, low autonomous reasoning ability, and insufficient multi task collaboration, significantly improving the production and service efficiency of robots and achieving more intelligent machine production, "said Zhao Gang.

　Zhu Keli, the founding director of the National Research Institute for New Economy, analyzed that in industrial production, embodied intelligence can solve two major challenges: replacing human labor in extreme working conditions and achieving micro level manufacturing accuracy. For example, oil platform inspection robots can replace manual entry into flammable and explosive areas to complete pipeline inspection. In the field of people's livelihood services, embodied intelligence focuses on resource optimization in high-frequency demand scenarios, such as supermarket replenishment robots automatically replenishing shelf goods at night.

　　Pu Songtao stated that overall, the product form and business model in the field of embodied intelligence are not yet mature enough, and the value of large-scale applications has not been fully demonstrated. However, in terms of development potential, the application value of embodied intelligence in improving production efficiency, enhancing job safety, and improving service convenience is immeasurable. Its greater economic and social value needs to be continuously explored and expanded in future applications.

Break through the core bottleneck

　　Zhao Gang believes that currently embodied intelligence needs to overcome three core technological bottlenecks. One is to break through the technical bottleneck of basic models and gradually build end-to-end visual language action models and world models based on existing large language models, greatly improving the perception, reasoning, prediction, and action capabilities of embodied agents towards the physical world; The second is to break through the performance bottleneck of key components of the body, such as inference chips, high-end harmonic reducers, six dimensional torque sensors, multimodal sensors, hollow cup motors, etc; The third is to break through the bottleneck of high-quality real datasets and provide more real environment data for model pre training.

　　The next 3 to 5 years will be a critical period for the large-scale development of embodied intelligence, "said Pu Songtao. From a technological perspective, intelligence capabilities will be further enhanced; From an industrial perspective, the diversification of machine forms and the high-end development of key components will evolve in parallel, resulting in a more diverse range of product forms. However, competition for key components will become more intense; From an application perspective, the exploration of applications in multiple fields and scenarios will continue. Finding effective application scenarios at certain stages and forming feasible business models is the direction of joint efforts in the industry.

　Zhu Keli analyzed that embodied intelligence will make significant progress in scene penetration, technological architecture, and industrial form in the future. Scene penetration has extended from "dangerous repetition" to "life companionship", covering the vast majority of high-risk positions in the industrial sector, and the penetration rate of home service robots has exceeded 30%; The technological architecture is evolving from "modularity" to "integration", with end-to-end large model driven deep integration of decision-making and execution; The industrial form is upgrading from "single machine intelligence" to "cluster collaboration", and the completion of cross scenario collaboration by a hundred robot formations will become the norm.

　　Scenarios play an important driving role in the large-scale application of new technologies and products. It is reported that the Ministry of Industry and Information Technology will promote the entry of industrial robots and humanoid robots into factories, with priority given to their practical application in segmented scenarios such as welding, assembly, spraying, and handling. Focus on promoting robot products in complex and harsh production environments such as mining, civil explosives, and emergency response, and enhance the level of intelligent operations in dangerous and harsh environments.