Launch of OPS-SAT satellite with embedded AI solutions from CIAR project
Activeeon tools accelerate the development and accuracy of AI on-board satellite launched Dec. 18th.
A “flying laboratory”, ESA’s OPS-SAT satellite has been launched into space as part of CHEOPS mission on a Soyuz-Fregat rocket from Europe’s Spaceport in Kourou, French Guiana
Since 2 years, Activeeon has been part of IRT Saint-Exupéry’s (Sophia-Antipolis) CIAR (Chaine Image Autonome et Réactive) project with the goal to implement neural network algorithms on OPS-SAT FPGA for the detection and classification of objects on satellite images.
On December 18th, the launcher Soyouz-Fregat successfully carried into space three ‘CubeSats’, small satellites based on standardised 10 cm cubic units, including ESA’s OPS-SAT – the world’s first free-for-use, in-orbit testbed for new software, applications and techniques in satellite control.
OPS-SAT’s purpose is testing and validating new techniques in mission control and on-board satellite systems. OPS-SAT is devoted to demonstrating drastically improved mission control capabilities, that will arise when satellites can fly more powerful on-board computers. The satellite is only 30cm high, but it contains an experimental computer ten times more powerful than any current ESA spacecraft.
Ops-Sat (photo credit ESA)
The heart of the OPS-SAT satellite payload is the processing platform, which is responsible for providing a reconfigurable environment able to fulfill the objectives of each experiment. The processing platform runs Linux, as the operating system consists of a flexible and reconfigurable framework, featuring sophisticated processing capabilities, interfaces, memory integrity and reconfigurable logic.
The platform consists of an ‘Altera Cyclone V SoC’ with an ARM dual-core Cortex-A9 MPCore and a Cyclone V FPGA.
Regularly, the code embedded on OPS-SAT will be updated from Earth.
There are many constraints such as memory consumption, electricity, execution speed, etc.
The role of ProActive software from Activeeon in this project is to help with the choice of hyperparameters or models to determine which neural networks or other algorithms to use on satellite image detection for the next code to load on OPS-SAT. The further mission of CIAR project is to:
- find the best hyperparameters of embedded algorithms in order to have a fast, high-performance algorithm, minimizing detection errors
- find the best implementation and hardware target for embedded algorithms : automate and parallelize code compilation for flashing different cards: FPGA, ASICs, TPUs,… for comparison
- facilitate data retrieval from various platforms (Copernicus, PEPS,…) for training of machine learning algorithms
These principles and techniques are also applicable in other fields, drones for example.The field of embedded systems is vast, and OPS-SAT is only one use case among others.
Learn more about OPS-SAT https://www.esa.int/Enabling_Support/Operations/OPS-SAT
December 18, 2019