What? Even Robots Need to "Go to School"? That's right. Just like humans, humanoid robots also need to "go to school" before they can start working. On January 8th, at the Hefei Embodied Intelligent Robot Data Collection and Training Ground in the Yangtze River Delta Digital Technology Demonstration Park in Yaohai District, more than ten robot "students" arrived. Here, they will learn skills such as cooking, making beds, and serving as shopping guides.

At 9 a.m. that day, the robot "classmates" were already immersed in their learning. In this 700-square-meter "classroom," each robot focused on its own tasks. Some were following instructions to pick up goods and place them in shopping carts, some were practicing how to spread bed sheets and organize wardrobes, while others were stir-frying with woks… Despite their seemingly proficient movements, they still rely on "teachers"—data collectors who guide them step-by-step.

"Some actions may seem simple, but for robots, they require a process of repeated practice. We need to guide the robots through exercises using teleoperation devices, and the various data generated during training will be captured and recorded by cameras to form a dataset for optimizing models," said Wu Meiling, a data collector at the training ground. She jokingly compared the process to helping robots "practice with exercises," with the main goal being to feed tens of thousands of pieces of data into the robot's "brain" so that it can learn to apply knowledge flexibly and adapt to real-world production and life scenarios.

It is understood that this embodied intelligent data training ground, jointly established by the International Center for Advanced Technology Application Promotion (Hefei) and Yaohai District, covers fully simulated scenarios in multiple fields such as industrial manufacturing, commercial services, and household services. It meets the closed-loop requirements for robots, from "product production and testing to training, data collection, and product iteration," while also providing effective data parameters to support the formulation of national standards for robot products and the industry.

"Previously, we replicated real manufacturing production lines for industrial scenarios, featuring leading enterprises such as LCFC and JAC. This recent upgrade and transformation focus on adding scenarios to tackle practical challenges in household services," explained Wang Zhiwei, Project Manager of the Robotics Division at Hefei Guoxian Holding Co., Ltd. He noted that the training ground constructs a 1:1 real home environment based on enterprise needs, providing diversified and high-quality datasets through hands-on training. This approach helps shorten research and development cycles, reduce trial-and-error costs, and offers replicable and verifiable scenario implementation solutions for the industry. Currently, the training ground has served six robotics companies. By March this year, it is expected to expand to 5,000 square meters and launch training for over 30 specialized scenarios.

Amid the wave of the new technological revolution and industrial transformation, embodied intelligence has become a new global competitive frontier. Since last year, a number of humanoid robot training grounds have been established across the country. Shanghai focuses on "heterogeneous" humanoid robots, Qingdao in Shandong emphasizes specialized scenarios such as marine applications, and Wuhan in Hubei prioritizes industrial closed-loop development. Hefei in Anhui, on the other hand, emphasizes scenario-based training and solutions, striving to become a technology transformation hub for manufacturing in the Yangtze River Delta region. As the humanoid robot industry enters a phase of rapid development, it is believed that more and more humanoid robots will "graduate" from here in the future, benefiting countless households and empowering a wide range of industries.

Source: anhuinews.com