The space sector is currently undergoing a technological disruption featured by onboard intelligence. By processing the information captured from the space directly onboard the spacecrafts with advanced AI techniques, onboard intelligence has the potential to realize more efficient, agile, autonomous, and reconfigurable Earth Observation (EO) systems.

In EO, the end users often expect acquiring valuable insights and making optimal decisions at low latency. Accordingly, considerable research efforts have been devoted into investigating onboard intelligence for EO applications like the early detection of natural disasters, vessel incidents, and gas leaks. Onboard intelligence also allows for identifying low-quality EO data, e.g., cloud-covered satellite images and remote sensing images containing limited information of interest, and then discarding them to save downlink data bandwidth to the Ground Station.

AI algorithms supported by dedicated hardware/software codesign can completely change the future of the EO missions by reducing developing time, costs, and human resources. IngeniArs has a long experience in space-FPGA designs and AI algorithms development, starting the prototyping of its qualified hardware/software codesign for AI applications a couple of years in advance.

The result is the GPU@SAT ecosystem.

It is composed of a hardware IP-Core for FPGA completely independent of the target technology, and a software framework that allows to program and configure the device. The flexibility and modularity of the GPU@SAT ecosystem are perfect when dealing with space requirements. In particular, its technology independence allows to implement it directly on space-grade FPGA. To simplify the development of the sub-activities, GPU@SAT ecosystem will be implemented on a representative device, resulting in a dedicated development kit: GPU@SAT devkit.