German satellite will use AI to detect anomalies on asteroids and planets

Key Takeaways:

– A German satellite called SONATE-2, designed and built by a team led by Professor Hakan Kayal, will test new AI technologies in orbit for automatic detection of anomalies on planets and asteroids.
– The satellite is a six-unit cubesat, a type of nanosatellite that is no bigger than a shoebox.
– The unique aspect of the mission is that the AI will be trained on board the satellite, rather than on Earth with powerful computers.
– This is important because it allows the satellite to investigate unknown objects in the solar system without needing to send data back to Earth for training.
– The satellite features four cameras that will provide images directly to help train the AI.
– The mission will also test other small satellite technologies, including a system for the automatic detection and recording of lightning and an electric propulsion system.
– SONATE-2 will launch aboard a SpaceX rocket in March 2024 and is expected to be operational for at least one year.
– The project is funded by Germany’s Federal Ministry of Economic Affairs with €2.6mn.
– Launches of nanosatellites for testing new technologies have been gaining traction, with other examples including Open Cosmos’ Menut and Estonia’s ESTCube-2.

The Next Web:

A German satellite that will test new AI technologies in orbit for automatic detection of anomalies on planets and asteroids is set for launch.

Despite its ambitious mission, the so-called SONATE-2 is a six-unit cubesat, a type of nanosatellite that’s no bigger than a shoebox. It was designed and built by a team led by aerospace engineer Professor Hakan Kayal from Julius-Maximilians-Universität (JMU) Würzburg in Germany.

According to Kayal, projects of this sort are quite uncommon. “What is unique about our mission is that the AI is trained on board. Normally, this training is done on Earth with powerful computers,” he said.

“Let’s assume that a small satellite is to investigate a new asteroid in the solar system in the future,” Kayal explained. “It cannot be trained for this task on the ground, because the object of investigation is largely unknown.”

In typical scenarios, this would mean that data collected from space would need to be sent back to Earth and then be used to train the AI remotely — which is a lengthy process for long-distance missions. But a higher level of autonomy supported by AI on board would be much more efficient.