Hesai Technology, a prominent Lidar technology manufacturer, unveiled its innovative 3D data engine, Kosmo, on October 14. This system includes the ultra-high-definition AI spatial camera Kosmo One and the Kosmo World cloud platform, designed to transform real environments into 3D digital spaces, thereby facilitating robot training. This initiative represents a significant shift in the company’s approach, which had until now primarily focused on sensor delivery, now expanding its business model towards data production and cloud services.
Kosmo consists of two key elements: the Kosmo One camera, which combines a high-resolution Lidar sensor with an image sensor and an inertial measurement unit, and the Kosmo World platform, which enables data reconstruction in 3D environments. This advancement is crucial, as robots, unlike autonomous vehicles, are not yet widely deployed, limiting the accumulation of data necessary for their training. Li Yifan, co-founder and CEO of Hesai, emphasized that Kosmo aims to convert real environments into accessible and modifiable training scenarios, significantly reducing preparation costs.
At the presentation, Hesai also announced strategic partnerships with artificial intelligence companies such as Galbot and Dexmal, and showcased collaborative results with Pixar and Amap, covering areas such as film production and cultural heritage preservation. Kosmo faces what Xiang Shaoqing, co-founder and CTO of Hesai, called the “impossible triangle” of embodied artificial intelligence, where achieving quality, scale, and low cost simultaneously is a challenge.
Kosmo’s efficiency is remarkable; an operator can scan a 100-square-meter space in just 15 minutes, generating more than 10,000 training trajectories from a single capture. This translates into a 99% success rate in real-world applications, as demonstrated in a use case at a Dairy Queen ice cream shop, where modeling time was drastically reduced.
Furthermore, Hesai plans to offer Kosmo to individual consumers in the future, expanding its user base beyond corporate R&D teams. This move towards spatial data services is part of a broader trend in the Lidar industry, where companies like RoboSense are also exploring spatial cameras. The urgent demand for data infrastructure in the embodied AI industry is driving this transition, aiming to standardize and optimize scene production for robot training. Kosmo’s ability to digitize the real world has significant implications not only for robotics but also for film production and cultural heritage preservation.