RM新时代有限公司

Photo from People's Daily (May 5, 2026, Page 2)

On the occasion of Youth Day, General Secretary Xi Jinping replied to representatives of China Youth May Fourth Medal and New Era Youth Pioneer Award winners, pointing out that this year marks thestart of the "15th Five-Year Plan," and it is the right time for young people to make contributions.

During the "May Day" holiday, at the Guangdong Embodied Intelligence Training Ground in Haizhu District, Guangzhou, various robots were undergoing immersive training under the operation of young architects and engineers: robotic arms precisely grasping car bolts for sorting, inspection robots autonomously patrolling along circuit pipelines to identify hazards, humanoid robots performing fine operations... On the front line of technological innovation, Chinese youth in the new era are writing a regretless youth with practical actions.

"Our task today is to simulate a supermarket scenario, having the W1 robot pick up and place common bottled drinks." Deng Yueci, a post-95s basic engine architect, said toCloudWalk TechnologySong Lei, a post-90s robot remote control engineer.

Photo of Song Lei controlling the robot to grasp objects. Hong Qiuting photographed

Taking a deep breath, pressing buttons to vibrate, moving his arms. Song Lei received the task, put on aVR (virtual reality) headset, and began operation.

As Song Lei moved his arms,the W1 robot's arms also swung.

"Speed shouldn't be too fast, keep the trajectory smooth." Deng Yueci reminded. Song Lei manipulated the VR controller, making the robot slowly approach a water bottle. The robotic arm paused slightly upon contact, and within one second, the robot perceived the interaction state with the object and recorded relevant data.

In Song Lei's view, there are many types of robots, but they are all as familiar and tacit as friends."When operating, imagine yourself as the robot to achieve a state of 'human-machine unity,' keep your muscles relaxed, and think about the details of the movements."

Deng Yueci stared at the sensors, closely monitoring the movement data of the robotic arm's end effector and joints."The degree of control is important. Picking up bottles of different materials requires slight differences in speed and force." Deng Yueci said, the more precise the movements and the more accurate the coordination with the robot, the more accurate and comprehensive the data collection.

After the remote operation, the collected dataset was transmitted to the server."The accuracy and comprehensiveness of the data need to meet high standards. For example, in the retail supermarket scenario being tested, the robot needs to be trained to recognize bottles of different materials and various types of shelves, and collect data separately." Deng Yueci said, a single collection is not enough; data must be augmented through simulation to generate new data. When the data reaches over 100,000 samples, the next step—model training—can begin.

During automated data collection, Song Lei noticed that the data collected by the robot was in raw format, which was inconvenient to read and use. How to optimize? Deng Yueci suggested rewriting a program to convert the raw format into a common format, improving the effectiveness of data collected by the robot in real operational scenarios.

From meticulously collecting data repeatedly to using simulated data to form models, and then deploying to real machines, the work is both hard and interesting."It's a bit like playing a 'level-breaking' game, with many points to explore." Deng Yueci told reporters, "As robots advance and grow in diverse scenarios, we are also growing."

When first controlling the robot, Song Lei's hands trembled uncontrollably, and wearing the headset made him dizzy. He often had to spend a long time getting familiar with the site."Carefully, start with the simplest actions, like making the robot grab a box, then try new moves." Now, whether it's a two-finger gripper wheeled robot in retail scenarios or a two-finger gripper and dexterous hand robot, Song Lei can operate them proficiently within 10 to 30 minutes. Even the large wheeled inspection robot used in complex workshops, which demands high balance from the operator, is no challenge for Song Lei.

Nevertheless, whenever a new robot is delivered to the training ground and unpacked for training, Song Lei still positions himself on the same level as the robot—a "newcomer" on the training ground. "What tasks can this robot actually perform? How can I make it do them well?" With these questions in mind, he practices daily, summarizes experiences, and continuously improves his operational skills.

"The development of embodied intelligence technology requires integrating multi-modal data across numerous scenarios," said Ding Ning, Director of the Guangdong Provincial Embodied Intelligent Robot Innovation Center. The training ground serves as both a laboratory and a training field, testing operational capabilities with real-world scenario demands, laying the foundation for robots to enter frontline production and daily life in the future. The young people on the training ground are playing a vital role.

"With lofty ideals and a commitment to hard work, these young people repeatedly train and experiment on the training ground, enabling robots to transition from 'specialized tools' to 'general-purpose intelligent agents,' thus contributing to the development of future industries," said Gao Fang, Party Branch Secretary and Chairman of the training ground operator, Guangdong Embodied Intelligent Technology Co., Ltd. Among the company's tech talents, those born after 1990 account for 74%.

The vast number of young people integrate personal pursuits into the national development agenda, continuously creating new achievements in their respective positions, and contributing youthful energy on the new journey.