The PhiSat-1 is a small satellite which is the perfect illustration of the space stage we are going through. Far from the large classic communication satellites of several tons sent by large Ariane 5 type launchers, PhiSat-1 (or ɸ-sat-1 for the scientific-Hellenists) is a large box of 10 x 20 x 30 cm which was launched with 51 other minisatellites on September 3 by the European low cost launcher, Vega.
This “carpooling” of satellites makes it possible to set up simple space missions at low cost. Costs are also contained on the side of electronic components: dedicated to experiments of a few weeks / months, they rarely integrate the expensive “reinforced” chips of the traditional space world. Rather, it contains consumer components, admittedly not certified to last for years under permanent electromagnetic bombardment of space, but very powerful and inexpensive.
Among this panel of chips, PhiSat-1 integrates for the first time an “Artificial Intelligence” processor signed by Intel, the Movidius Myriad 2. It will indeed fulfill an “intelligence” mission in the service of the quality of the data flow. .
Improve the quality of the data to be processed
PhiSat-1 is ESA’s first AI-enhanced observation satellite. Its purpose is less to scan the earth’s surface (which it will do anyway) than to improve data flow. Because in space, data (images, measurements, etc.) are a big challenge.
Besides the fact that these low orbit satellites are constantly moving and therefore need to switch quickly from one antenna to another to retrieve data, the transfer speed is generally quite slow compared to optical fiber, for example. . Even though the images generated can quickly represent terabytes of data.
Thanks to its Movidius Myriad 2 chip (which is not even of the latest generation sinceIntel has already launched a Myriad X since) and algorithms trained by deep learning, PhiSat-1 will be able to learn to sort through the images it sends. It will thus avoid encumbering the limited bandwidth of the satellite with unusable images on the ground.
By testing these algorithms with this mini satellite, ESA will be able to improve its spatial data sorting models thanks to an Intel chip that costs only a few tens of dollars. An obvious mark of the democratization of access to space thanks to consumer components.