Artificial intelligence is everywhere in modern life, but it had not yet arrived in space and it is Intel who has taken this step.

On 2 September, an experimental satellite of the size comparable to a cereal box was launched by a rocket together with 45 other satellites of similar size. The satellite, called PhiSat-1, is now in heliosynchronous orbit at 530 km of altitude and at a speed of over 27,500 km/h.

PhiSat-1 is equipped with a new hyperspectral thermal chamber with data processing managed by artificial intelligence thanks to the Intel Movidius Myriad 2 VPU Features the same chip that is located inside many

PhiSat-1 is part of a pair of satellites whose mission is to monitor polar ice and soil humidity, as well as testing intersatellite communication systems with a view to creating a network of Confederate satellites.

What is the first problem Myriad 2 is helping to solve? The management of the large amount of data generated by high-fidelity cameras such as that used on PhiSat-1.

At any given moment about 2/3 of the planet’s surface are covered by clouds and this means that numerous useless images are taken, saved, then transmitted to the ground using valuable gang resources, again saved, finally evaluated, hours or days later, by a scientist (or

The idea developed by the team was to use the processor on board the satellite to identify and eliminate images obscured by clouds, saving about 30% of the bandwidth.

However, Myriad 2 was not designed to go into orbit. Normally, computers used in space aircraft use highly specialized radiation-proof chips that may be up to two decades older than the state of the art of commercially available technologies. That is why so far the AI did not fall within the equation.

The Ubotica team then performed a… radiation characterization activity, performing a series of tests on the Myriad chip to check how to handle any calculation errors or simply wear. Intel’s artificial intelligence product has been able to pass all the tests it has undergone without requiring any modification compared to the commercial version.

AI algorithms are normally built using large amounts of data from which to learn However, in this case it was a matter of learning to understand what is and what is not a cloud. The Lupin had to instruct its application using synthetic data extracted from existing missions.

All these integration and testing activities for systems and software, involving six organisations across Europe, took place over four months. Unfortunately, a series of unrelated events • delays with the vector rocket, the covid pandemic-19 and unfavorable summer winds • have resulted in a wait of over a year to verify that PhiSat-1 worked

The launch took place in French Guiana on September 2nd; from a first verification, the satellite saved all images and recorded any decision made by the AI not to record images of clouds, so that the ground team could verify that the system worked as planned

By sending only useful pixels, the satellite can now improve bandwidth usage and significantly reduce aggregate downlink costs, not to mention saving time for scientists on the ground.

In future, the possibilities for using low-cost and small-scale satellites enhanced by artificial intelligence will be countless, especially considering that multiple applications will be implemented.

For example, by flying over areas subject to forest fires, a satellite will be able to detect and inform the authorities in minutes rather than hours. By overtaking the oceans, normally ignored, a satellite will detect pirate ships or environmental incidents. Above forests and agricultural areas, the satellite can track soil moisture and crop growth. Above the ice it will be able to trace its thickness and melting areas, helping to monitor climate change.

Many of these possibilities will soon be tested: ESA and Ubotica are collaborating with PhiSat-2, which will bring another Myriad 2. PhiSat-2 will be able to run AI applications that can be developed, installed, validated and used on the spacecraft during flight through a simple user interface.

For Intel, the artificial intelligence market applied to space projects is still a small market, but the potential impact is out of the question.

Leave a Reply

Your email address will not be published.

You May Also Like