The increasing of the resolution of the imager sensors entails to an increasing of the data to transmit on ground. To compress such amounts of data dedicated algorithms are usually required, such as CCSDS123, CCSDS 122, etc. In fact, to efficiently run these algorithms onboard satellites, dedicated hardware implementations are adopted. Recent studies have proof that AI algorithms seem to reach similar compression rate but strongly relaxing the computational requirements.  

GPU@SAT devkit is the perfect candidate to implement compression algorithms based on AI. The activity proposes to develop a compression algorithm for GPU@SAT devkit and evaluating its performance with respect to the standard algorithm, deriving a full benchmark of possible network and potential technological gaps.