Increasing the sensors autonomy is becoming of paramount importance to improve reliability and forecast potential faults before they happen. These systems can be implemented both with traditional algorithms, for example, FDIR, watchdog, etc, or with innovative applications such as FDIR supported by AI.  

FDIR supported by AI strongly relies on satellite sensor data trends, learning their behaviours directly from the data. These data-guided models offer the possibility to interpolate multiple sensors at the same time, potentially forecasting faults several hours in advance. GPU@SAT devkit represents the heart of the future IngeniArs flight boards, indicating it as the perfect candidate for such applications. Therefore, the sub-activity aims to develop an AI-FDIR system dedicated to EO. This system will be finally benchmarked with state-of-the-art FDIR algorithms adopted on EO.