Empowering Liard! The country's largest humanoid robot training ground officially opened

Recently, the Beijing Shijingshan Humanoid Robot Data Training Center (Phase II), the largest humanoid robot data training ground in China, was officially put into operation. This industrial hub, led by the Shijingshan District government and jointly built by Beijing-Shi Science and Technology Innovation Group, Beijing Shijingshan Bank Insurance Industrial Park, Leju Robot and Liard Virtual Moving Point, is not only a benchmark practice in implementing the State Council's "Opinions on In-depth Implementation of Artificial Intelligence + Action" to strengthen the requirements for the construction of high-quality data sets, but also marks a key step in my country's embodied intelligence industry in the field of data resource reserve and scene-based training, and lays a solid foundation for accelerating the evolution of humanoid robots ' embodied brains and promoting its large-scale application in scenarios such as industrial manufacturing.​

Government-enterprise collaborative construction of hubs: solving the bottleneck of specific intelligent data

Breakthroughs in embodied intelligence are inseparable from the feeding of high-quality data, and the scarcity of scene-based data and inconsistent training standards have always been the core pain points that restrict the industrialization of humanoid robots. The opening of the Beijing Shijingshan Humanoid Robot Data Training Center (Phase II) is a precise layout aimed at this industry problem. The center has built a leading domestic intelligent data production and training platform by integrating multiple advantages such as government policy guidance, enterprise technology research and development, and park resource carrying.​

In terms of spatial layout, the center has created an immersive training scene of tens of thousands of square meters, restoring 16 sub-scenarios in four categories: industrial intelligent manufacturing, smart home, health care services and 5G integration in a 1:1 manner, from auto parts sorting to family escort assistance, from warehousing and logistics handling to intelligent inspection and navigation, achieving full coverage of application scenarios from industrial to consumer. This highly simulated scene setting allows humanoid robots to learn skills in a near-real environment, providing a guarantee for the authenticity and effectiveness of data collection.​

From the perspective of industrial value, the center will not only expand the reserve of high-quality data resources for the specific intelligence industry, but also standardize industry training data standards through standardized data production processes and enhance the region's industrial agglomeration effect in the field of specific intelligence. With the subsequent cross-regional linkage with training grounds in Suzhou, Jinan, Hefei, Zhengzhou and other places, a specific intelligent data hub covering the whole country will soon be formed. It is estimated that the monthly data production capacity will reach 5000 hours, and the annual output of more than 6 million high-quality data will fundamentally alleviate the dilemma of insufficient data supply in the industry.​

Technology empowerment reveals the core: OptiTrack becomes training infrastructure

In this giant robot training ground, the OptiTrack system of Liard Virtual Moving Points has become an indispensable technical core. It is like equipping robots with precise sensing and data recording centers, providing skills learning and data collection for humanoid robots. Provide full process support. As the world's leading spatial positioning equipment, the OptiTrack system has become an infrastructure-level facility in the intelligent industry chain with its core advantages of high precision, low latency, high frame rate, and large space coverage.​

Its technical advantages are fully released in training scenarios: through optical and marker-free fusion motion capture technology, relying on high-sensitive sensor arrays and deep learning algorithms, the system can accurately identify the robot's limb movements and even subtle joint changes, and record it in real time. Operating data and environmental interaction information under different scenarios. With the support of the OptiTrack system, the skill learning efficiency and execution accuracy of the humanoid robot in the field have been significantly improved. At present, it has mastered more than 20 atomic skills such as empty container return, material sorting, weighing and packaging, and product boxing. The execution success rate in various tasks exceeds 95%, and it can calmly perform diversified tasks such as handling, inspection, navigation, and distribution.​

This technical empowerment is not limited to Shijingshan Training Center. At present, the Liard·Virtual Moving Point OptiTrack system has been widely used in many industry organizations such as Futian Laboratory in Greater Bay Area, Ubixuan Robot Training Center, and Leju (Shenzhen) Robot Training Center. Its stable and reliable performance has been fully verified by the market and has become an important force in promoting the unification of industry technical standards.​

Three-dimensional layout for the long term: full chain support from data to application

This in-depth participation in the construction of Shijingshan Training Center is the epitome of the three-dimensional layout of Liard Virtual Moving Points in the field of embodied intelligence. Since joining the company, relying on its profound accumulation in the fields of AI and spatial computing, the company has built a three-in-one intelligent robot industry service system that integrates data-platform-application. It is committed to solving the problems of large-scale and industrialization of robots and promoting the industry froma single point. Technological breakthroughs move towards large-scale scenarios.​

At the data level, by participating in the construction of a national training center network, we will build standardized data production capabilities to provide robot companies with high-quality, scene-based training data services; at the platform level, with the OptiTrack system as the core, we will create a technical support platform integrating motion capture, data collection, and skills training to meet the training needs of different types of robots; at the application level, relying on a deep understanding of industry scenarios, we provide customized robot training and technical adaptation solutions in areas such as automobile manufacturing, logistics and warehousing, and health care services., accelerating the transformation of technical achievements into practical applications.​

The value of this full-chain layout is gradually emerging: when the robot completes skill learning in the training center, Liard Virtual Moving Point can help the company complete the adaptation from training ground skills to actual scene applications by relying on its grasp of the application scenarios. Optimize and shorten the technology implementation cycle. For example, in an automobile manufacturing scenario, centrally trained robots can quickly adapt to the material handling needs of the workshop assembly line, and reduce the company's labor costs and operational risks with high success rate operation performance.​

Looking to the future: The leap from training ground to industrial field

With the opening of the Beijing Shijingshan Humanoid Robot Data Training Center (Phase II) and the gradual formation of the national data center, my country's embodied intelligence industry is about to enter an accelerated development period of data-driven + scenario iteration. As a core technical participant, Liard Virtual Moving Point will continue to base itself on the advantages of spatial computing technology in the future and continue to exert efforts in three directions: first, deepen the technical iteration of the OptiTrack system and improve tracking accuracy and anti-interference in complex environments. Ability; second, promote collaborative operation of cross-regional data hubs to achieve efficient transfer and sharing of data resources; The third is to strengthen in-depth cooperation with robot companies and create customized training solutions for different industry scenarios to help overcome technical problems in the field of personalized intelligence"".​

In the long run, this customized intelligent layout promoted by government and enterprises and empowered by technology companies will not only accelerate the large-scale application of humanoid robots in automobile manufacturing, logistics and transportation, etc., but will also drive spatial computing, motion capture, The coordinated development of upstream and downstream industries such as artificial intelligence provides strong support for my country to seize the lead in global customized intelligence competition. When the training ground"