The aim of my research is to develop a low-power onboard data processing system based on edge AI, specifically designed for space environments or autonomous robotic platforms. As space missions move farther from Earth, real-time ground control is constrained, making onboard intelligent systems essential for autonomous decision-making, anomaly detection and rapid response to unforeseen conditions.
The proposed system will integrate machine learning and neural network algorithms capable of analyzing heterogeneous data streams, including environmental images acquired by cameras and chemical or biological signals obtained with electrochemical sensors. The core objective is to design and evaluate an AI-driven processing architecture that can autonomously interpret these data to support crew health, environmental safety and system reliability. The AI will classify contaminants, recognize visual and biochemical patterns and predict potential hazards, enabling early warnings in the event of abnormal conditions such as water contamination, system malfunctions or hazardous environmental features. For this purpose, I will investigate the suitability of different neural network architectures, including convolutional neural networks for image recognition tailored for real-time onboard processing.
A key aspect of this research is the implementation of AI algorithms on low-power microcontrollers (such as STM32), which represent the most viable solution for space missions characterized by strict energy constraints. Unlike high-performance GPUs commonly studied in literature, these platforms are optimized for efficient computation under limited power availability, making them suitable for long-duration missions. In parallel, I will study the effects of ionizing radiation, characteristic of the radiation harsh space environment, on both hardware components and AI models. By exposing processors and learning algorithms to X-rays, proton beams and mixed-field radiation environments, I will experimentally assess functional faults, performance degradation and inference errors. The goal is to classify radiation-induced effects and understand how they impact AI reliability, robustness and learning behavior in space conditions.
Compared to the current state of the art, this research adopts a unified approach that combines low-power edge AI, onboard autonomy and systematic radiation testing. By the end of the project, I aim to deliver a prototype system together with a comprehensive characterization of radiation effects on AI-based data processing, contributing to the development of more autonomous, resilient and energy-efficient space exploration technologies.

Click here to view my poster.