About Me

Hello! I'm Matteo, currently a PhD Student at the University of Luxembourg.

My research focus is on improving Reinforcement Learning (RL) algorithms in the context of autonomous complex robotic systems.

Previously, I worked for 3+ years as a Machine Learning Engineer at Argotec, an aerospace company specializing in micro-satellites for Deep Space. I contributed to the core algorithms used for the autonomous visual navigation of LICIACube and ArgoMoon, satellites that were part of NASA's DART and Artemis I Missions.

I have a Master’s in Computer Science and Engineering (AI track) from Politecnico di Milano and a double BS degree from PoliMi and Tongji University.

While building strong foundations in Machine Learning, complex systems dynamics, and related subjects through formal education, I like to pursue my interest in human behavior and psychology by reading about social dynamics, individual and collective intelligence. My drive is to build and understand AI technologies that can explore decision-making in ways novel to humans and potentially beneficial for tackling humanity's most pressing world problems.

Research & Projects

ICE2THRUST (2024 - present)

Solar-Electric Water Propulsion System

As research expert in the development of RL-based autonomous navigation systems for the ICE2THRUST project, I'm working on autonomous GNC strategies for satellites aimed at demonstrating the next-generation propulsion technologies that utilize water as propellant. This European EIC Pathfinder project goal is to demonstrate the world's first In-Situ Resource Utilisation (ISRU) end-to-end process chain.

Development of simulation environments for 3DoF and 6DoF spacecraft dynamics
Implementation of RL-based modular navigation systems for different parametrizable satellites
Sim-to-Real transfer techniques for RL algorithms in realistic laboratory settings

For more information, visit ice2thrust.space.

AIRSHIP (2023 - present)

Autonomous Guidance, Navigation, and Control for Unmanned WIG Vehicles

As Lead Research Scientist in the development of RL-based autonomous navigation systems, I'm working on autonomous Guidance, Navigation, and Control (GNC) strategies for innovative Unmanned WIG Vehicles (UWV). This Horizon Europe project focuses on developing cutting-edge autonomous systems for next-generation aviation.

Design and simulation of advanced GNC strategies for WIG vehicles
Deployment of RL-based policies for autonomous maneuvering with enhanced speed, flexibility, and energy efficiency
Integration of machine learning algorithms for real-time decision making in complex environments

For more information, check out airship-project.eu.

NavBench (2025)

A Unified Robotics Benchmark for Reinforcement Learning-Based Autonomous Navigation

NavBench extends IsaacLab with a modular and scalable design, enabling unified definitions of tasks and robots, and training one policy per robot-task pair using a shared infrastructure. It uniquely supports sim-to-real transfer across diverse platforms and mediums for evaluating navigation in varied environments.

For more information, visit the NavBench project website.

DRIFT (2024)

Deep Reinforcement Learning for Intelligent Floating Platforms Trajectories

DRIFT introduces a comprehensive deep reinforcement learning framework for controlling floating platforms in both simulated and real-world environments. These platforms serve as terrestrial analogs for microgravity conditions, enabling robust testing of space navigation systems with seamless sim-to-real transfer capabilities.

Key Innovations:

First RL-based control suite for floating platforms with proven sim-to-real transfer
Advanced uncertainty handling for dynamic and unpredictable environments
High-precision position and orientation control under external disturbances
Comprehensive evaluation framework for autonomous navigation systems

REACT (2021)

Reinforcement Learning for Underactuated Satellite Attitude Control

This research investigates the application of Deep Reinforcement Learning for satellite attitude control, addressing both nominal operations and critical underactuated scenarios. The work demonstrates robust control policies capable of handling actuator failures for nanosatellites.

Key Contributions:

Novel RL-based attitude controllers for nanosatellites using reaction wheels
Robust policies for underactuated scenarios with random actuator failures
Real-time performance validation in hardware-in-the-loop test-bed

Publications

NavBench: A Comprehensive Benchmark Suite for Spacecraft Autonomous Navigation -- Matteo El-Hariry*, Antoine Richard*, Ricard M. Castan*, Luis F. W. Batista*, Matthieu Geist, Cedric Pradalier, Miguel Olivares-Mendez (Under Review, 2025)
DRIFT: Deep Reinforcement Learning for Intelligent Floating Platforms Trajectories -- Matteo El-Hariry, Antoine Richard, Vivek Muralidharan, Matthieu Geist, Miguel Olivares-Mendez -- IROS '24 (oral presentation) & MASSpace'24 (International Workshop on Autonomous Agents and Multi-Agent Systems for Space Applications)
Highly-Parallelised Simulator for Reinforcement Learning based Autonomous Navigating Spacecrafts -- Matteo El-Hariry, Antoine Richard, Miguel Olivares-Mendez -- 17th Symposium on Advanced Space Technologies in Robotics and Automation, 2023
Mobility Strategy of Multi-Limbed Climbing Robots for Asteroid Exploration -- Warley FR Ribeiro, Kentaro Uno, Masazumi Imai, Koki Murase, Barış Can Yalçın, Matteo El Hariry, Miguel A Olivares-Mendez, Kazuya Yoshida -- Climbing and Walking Robots Conference, 2023
Deep Reinforcement Learning Policies for Underactuated Satellite Attitude Control -- Matteo El Hariry, Andrea Cini, Giacomo Mellone, Alessandro Balossino -- NeurIPS 2021 Workshop on Deployable Decision Making in Embodied Systems
A Validation and Calibration Methodology for Autonomous Navigation of Small Satellites -- Sarah Ciaglia, Nicolò Roberto Benigno, Matteo El Hariry, Pasquale Tricarico, Valentina Marchese, Niccolò Battezzati, Simone Simonetti -- IAC (International Astronautical Congress) 2021