As part of the ASTRA research group at Politecnico di Milano, my doctoral research focuses on advancing autonomous Guidance, Navigation, and Control (GNC) systems for planetary landing. In recent years, securing reliable access to planetary surfaces has emerged as a cornerstone of international space strategies. This access is foundational for enabling In-Situ Resource Utilization (ISRU), a critical requirement for sustained deep-space exploration that demands precise and safe touchdown capabilities in increasingly complex environments.

Current vision-based autonomous GNC systems have demonstrated substantial operational value in real missions, significantly improving landing site accuracy and Hazard Detection and Avoidance (HDA). However, these systems possess intrinsic limitations. They predominantly rely on sensors operating exclusively within the visible light spectrum, restricting their application to well-illuminated scenarios. This dependency makes reaching high-value scientific targets located in permanently shadowed regions—such as the Lunar South Pole—highly complex, if not entirely unfeasible. Furthermore, current architectures have yet to fully integrate the capabilities of Artificial Intelligence (AI) to optimize navigation processes.

To address this operational gap, my research aims to develop and validate a robust optical navigation filter that integrates classical computer vision algorithms with AI. Crucially, this system processes multispectral data, fusing the visible spectrum with Thermal Infrared (TIR). Specifically, the implementation of Long-Wave Infrared (LWIR) microbolometers offers a highly robust, low-complexity solution. These sensors provide an optimal sensitivity range for the extreme temperature gradients found on planetary surfaces, and could potentially ensure that the GNC system remains completely independent of solar illumination conditions.

Currently, my methodological approach focuses on numerical modelling, driving the development of a high-fidelity image generator capable of rendering physically accurate planetary surfaces (primarily targeting Lunar and Martian scenarios) across both visible and TIR spectra. Concurrently, I am optimizing image processing architectures to handle this multispectral data. A key focus is evaluating Convolutional Neural Networks (CNNs) alongside classical methods to identify the most effective solutions for both robust feature extraction and sensor fusion.

To support the empirical validation of this research, I am also contributing to the design of a laboratory infrastructure dedicated to testing vision-based GNC algorithms. During the second phase of my doctoral studies, this facility will be utilized to transition the filter from numerical simulation to Hardware-in-the-Loop (HIL) testing. The ultimate objective and my primary contribution to the "Space It Up!" project is the delivery of a HIL-validated, multispectral optical navigation filter capable of enabling autonomous precision landings in unlit environments, directly facilitating future ISRU operations.

Click here to view my poster.