I completed my PhD in Aerospace Science and Technology at the University of Bologna, Italy, under the supervision of Professors Paolo Tortora and Dario Modenini. I began the programme in November 2021, right after apost-graduate research scholarship,and defended in January 2026. The thesis is titled "Deep Learning for Spacecraft Pose Estimation: From Algorithms to Hardware-in-the-Loop Testbed".

My doctoral studies are fully supported by a research scholarship on "Autonomous navigation of satellites for proximity and docking operations" at the Microsatellite and Space Microsystems Laboratory at Forlì Campus.

During my PhD journey, I have taken on roles as a teaching tutor and MATLAB Ambassador. Additionally, I have actively participated in various laboratory activities, includingstudent supervision,involvement in#T-TeC 2021,and the contribution to the development of theNanoDynA ADCS facility for ESA-ESTEC.

From late February to mid-July 2024 I was a visiting researcher at theSentient Satellites Labof the University of Adelaide (Australia), working on lightweight detection of unknown satellites and structure from motion.

The objective of my PhD is to contribute to the advancement of autonomous vision-based navigation techniques. This involves leveraging lightweight and cost-effective sensors, particularly monocular cameras, within the context of close-proximity and docking grasping operations. Additionally, it involves the application of Deep Learning for image processing.

In recent years, In-Orbit Servicing (IOS) and Active Debris Removal (ADR) missions have gained significant attention in the space community. These missions aim to inspect, recover, refuel, relocate, upgrade, and deorbit unresponsive satellites or assets approaching the end of their operational life. This marks a transformative shift in satellite design philosophy while ensuring safe and continuous access to space. These missions involve at least two probes maneuvering in close proximity, potentially leading to docking.

Close Proximity Operations (CPOs) in such scenarios pose a high risk due to the possibility of impacts, particularly when dealing with noncooperative targets. Despite ongoing discussions within international working groups, there is currently no agreed-upon international standard for safe CPOs. However, design guidelines and best practices are being assessed. These guidelines emphasize that the chaser must be able to measure and control its state relative to the client, ensuring adherence to a defined approach corridor during the rendezvous.

In the initial phase of my PhD, I focused on refining a pose estimation software developed in a prior research scholarship. This refinement process extended to testing the software on a Coral Dev Board Mini. Subsequently, I delved into the 2nd ESA Pose Estimation Challenge, where my attention shifted towards addressing domain gap issues. Concurrently, I explored enhancing the robustness of my pose estimation pipeline by incorporating refinement steps into both the detection and pose regression processes. Concerning the domain gap, I actively worked on minimizing disparities between real and synthetic images using generative adversarial models, a strategy that proved effective for bridging the gap.

During my research period at the Australian Institute for Machine Learning, University of Adelaide, under Professor Tat-Jun Chin, head of the Sentient Satellites Lab, I worked on:

• Lightweight detection of unknown satellites (PoseBowl challenge).
• Structure from motion for reconstructing uncooperative targets.

In the final part of the PhD I designed and commissioned a 7-degree-of-freedom robotic hardware-in-the-loop testbed to reproduce relative camera-target motion under solar-spectrum illumination, for generating labelled real images and, in future work, testing pose estimation algorithms. Details on the testbed page.

📰 For information regarding publications, please refer to the dedicated section.