Satellite sensors collect an incredible amount of data, but the limitations of on-orbit computing prevent operators from processing it in space. Aethero, founded a year ago, plans to change that.
Aethero develops radiation-protected edge computers for on-orbit data processing and, ultimately, autonomous decision-making.
«Space data pre-processing is currently a larger market, but we expect that over the years and as the number of orbital assets increases, spacecraft autonomy will become mainstream,» said Aethero co-founder and CEO Edward Ge in a recent interview.
Ge co-founded Aethero with his childhood friend Amit Pinnamaneni, and they also own Stratodyne, a startup that creates high-altitude balloons for remote sensing. Although these are quite different areas, the founders say that both businesses face similar challenges.
Space computers today use field-programmable gate arrays (FPGAs) with older architectures that cannot handle computationally intensive tasks such as training or deploying advanced computer vision models on orbit.
«We realized that the problem is not getting enough data from space, but getting the data from the sensor in space to the end user fast enough, and enabling satellites in orbit to make real-time decisions».
During the previous round of funding, Aethero raised $1.7 million.
The first generation of the startup’s space computer is called AetherNxN and is based on the Nvidia Orin processor. It is designed for 7-10 years of life in low Earth orbit and can fit on small platforms such as CubeSats. It provides 20 times more computing power than existing options, Ge says.
Aethero plans to release a larger second-generation computer before switching to its own space processor. According to the company, this will help ensure lower power consumption and high performance. The startup aims to produce processors together with Intel by about 2026.
One of Aethero’s missions will launch in June this year on the SpaceX-11 transporter to demonstrate the ability to work in space, deliver updates by air for onboard computer vision models, and train computer vision models.
«We see ourselves as an Intel or Nvidia in the space industry,» Ge said.
Source: TechCrunch