2026年服贸会期间,北京市石景山区具身智能训练中心完成一期升级。该中心位于首钢园,重点布局4D高斯光场与世界模型技术,旨在通过高精度数据采集,为具身智能产业提供硬件与数据底座支撑。
AI-generated summary
石景山区具身智能训练中心位于北京首钢园,由多家产业链企业共同建设。该中心旨在通过数据采集与模型训练,推动具身智能产业发展。
中新网北京9月12日电 (记者 陈杭)记者12日从北京市石景山区获悉,2026年服贸会期间,北京市石景山区人形机器人数据训练中心一期正式升级,以石景山具身智能训练中心的全新身份亮相。
训练中心坐落于北京首钢园科幻产业创新中心,占地面积1500余平方米,由凌云光、中关村通力牵头,京石科创、影身智能、天链机器人等多家产业链企业共同参与建设。设置4D世界采集区、物理交互实验场、通用小脑基座数据采集区、多场景4D共享采集区、AI工厂等功能区域,以数据生产环、智能进化环两大闭环串联整体空间。
训练中心本次升级重点布局4D高斯光场与世界模型两大核心技术路线,在原有二维视频、动作轨迹数据基础上,叠加深度、空间关系与物理信息维度,完整刻画物体位姿、接触事件、力学关系与任务状态演进,支撑模型更好理解“动作如何改变世界”。
在数据采集能力上,训练中心新建多视点光场采集系统,能高精度还原物体运动轨迹、形变过程及光影变化,产出高保真原始训练素材。面向家庭场景,系统可完整复现家务清洁、烹饪备餐、家庭照护等真实动作,输出高质量家庭场景数据集;面向工业制造领域,深度采集布料缝纫、打包搬运、产品装配等作业流程,夯实机器人对工业物理规律的认知基础。
目前,北京市石景山区具身智能训练中心一期已完成升级,二期、三期已有效运营。下一步,将推动一二三期能力贯通协同,整体打造集真机数据采集、真人数据、4D光场重建与世界模型仿真推演于一体的综合性具身智能基础设施,为国内具身智能产业发展提供关键硬件与数据底座支撑。
AI outlook — possibilities, not facts
推动一二三期能力贯通协同
Likely · Within months
The Liuzhou UBTECH industrial humanoid robot super smart factory was officially put into operation on the 12th, becoming the first factory in Guangxi to realize the closed-loop "making robots with robots". The base mainly produces Walker S and Cruzr series robots, with an annual planned production capacity of over 10,000 units, aiming to create an innovation demonstration highland for humanoid robots for ASEAN.
The 2026 China Computing Power Conference opened in Langfang, Hebei, and the country's first national computing power interconnection area (Hebei) node was officially launched. This node is implemented by China Mobile Hebei Company and will enhance the computing power hub function of Xiongan New Area and promote the cross-regional circulation and unified dispatch of computing power resources in Beijing, Tianjin and Hebei.
On September 12, the 2026 China Computing Power Conference Xiongan Forum was held in Langfang. At the meeting, the Xiongan International Computing Power Integrated Dispatch Center was launched. The center consists of a dispatching platform and a large model service platform, aiming to build a national computing power hub through cross-domain computing power dispatching and large model aggregation services.

Wu Jinsong, a professor at Fudan University, said in Shanghai that brain-computer interface is a key port for the interaction between humans and artificial intelligence. He emphasized that industrial development should be oriented towards clinical issues and realize technology implementation through the intersection of medicine and engineering, and pointed out that in the future, neurotechnology ethics and safety governance need to be simultaneously strengthened.

According to the analysis of MIC, the AI data center competition has shifted from hardware to interconnection and system layers. As the scale expands, communication networks have become the key to releasing computing power. Taiwan's supply chain needs to shift from simply supplying hardware to developing integrated capabilities of "connecting computing power, adjusting computing power, and maintaining computing power" to seize the next stage of business opportunities.

The 2026 first Zhongguancun Science City Photonics Industry Ecological Conference was held in Beijing. Experts emphasized the importance of photonic technology as a strategic force, and pointed out that through the "Innovation Cup" competition and the construction of a common technology platform, it aims to open up the industry-university-research chain and promote photonic technology from the laboratory to the production line.