State‑level Embodied AI Platform Lands in Chengdu to Train Robots with Sourced Datasets

On July 11, the Chengdu working group of the OpenAtom OpenLET embodied intelligence open‑source dataset community was inaugurated at the University of Electronic Science and Technology of China (UESTC).

Embodied intelligence is a brand‑new growth sector in artificial intelligence, following closely after large models. However, unlike large models that “feed” on massive amounts of internet text, embodied intelligence requires robots to repeatedly trial, act, and perceive in the real physical world in order to accumulate effective data. Yet solo data collection with real robots is extremely costly and inefficient; the shortage of data and developers has become the industry’s biggest bottleneck.

As a national‑level open‑source dataset community for embodied intelligence, OpenLET’s landing in Chengdu will leverage the city’s industrial foundations in intelligent manufacturing, electronic information, and robotics, as well as integrate Sichuan’s high‑quality university research resources and advantages in the information technology sector. This will significantly lower technical R&D barriers and help more developers easily engage in robotics and embodied intelligence application development.

“We have already open‑sourced over 60,000 minutes of real‑robot datasets,” said Wang Song, Director of the OpenLET Technical Committee. These valuable data—derived from actual robot walking, grasping, obstacle avoidance, and interaction—enable ordinary developers to start robot development at a low threshold. “Users no longer need to buy expensive robots and conduct repeated lab experiments; they can directly call the open data to train algorithms and validate ideas,” he added.

Cheng Hong, Dean of the School of Mechanical and Electrical Engineering at UESTC and Director of the OpenLET Chengdu Working Group, introduced that the working group will collaborate with universities and enterprises in Sichuan to supplement OpenLET’s existing datasets, forming and open‑sourcing a southwestern region‑specific vertical‑domain data repository. This will reduce data processing costs for universities and SMEs.

Specifically, on one hand, it will engage in deep cooperation with more than 40 universities in Sichuan, including Sichuan University, UESTC, and Southwest Jiaotong University, focusing on real‑robot data sharing, joint technical research, and professional talent cultivation. On the other hand, it will collaborate with research institutes and robotics companies in the Chengdu‑Chongqing economic circle to help western China seize opportunities in the humanoid robotics industry. “Our goal is simple: to enable embodied intelligence to empower Sichuan’s real‑economy sectors such as intelligent manufacturing and public services,” said Cheng Hong.

Original link: https://cbgc.scol.com.cn/news/7751448?0=