A robotic arm delicately picks up a cup, another precisely places ingredients for a beverage—these are just some of the tasks humanoid robots perform at the Second World Humanoid Robot Sports Games, feats that rely on vast amounts of real-world training data. But where does this crucial data come from?

On September 3rd, reporters from the "Vitality China Research Tour" Jiangsu leg visited the JD.com Embodied Intelligence Data Collection Center in Suqian, a facility serving as a data "granary" for training robots in everyday human skills. The center, described as the first of its kind for robot data collection in Jiangsu, is where raw information from countless real-life scenarios is gathered to teach machines how to interact smoothly with their surroundings.

During the visit, reporters observed how the center captures and processes real-world interactions to train robots in tasks like drink preparation and item sorting. The insights gained here reveal the complex process behind creating robots capable of seamless, human-like responses—a development that promises to bring embodied intelligence closer to everyday life.