In the latest years, an incredible number of micro- and nano-satellites have been launched. To maintain a high production rate, more and more companies adopt COTS devices. These devices are designed for the commercial market, and they do not meet the requirements, reliability and qualification needed for a space device. In addition, connecting these COTS devices to satellite sensors available onboard represent a big integration challenge, i.e., different standard connectors, thermo-mechanical design, dedicated dissipation systems, etc.

GPU@SAT ecosystem combines the flexibility and the qualification to be adopted in any type of mission. Thus, such system strongly benefits of an easy method for managing new sensors at applications level without the need to redesign the entire software every time.

Therefore, this sub-activity aims to develop a virtualization layer on top of the GPU@SAT devkit, which shall reduce the effort of managing and elaborate data acquired by different imagers through innovative application paradigms: AI, CV, HPC. In addition, the virtualised layer shall exploit the hardware accelerator available on the GPU@SAT devkit.