Arthur Louette
PhD Researcher · Co-Founder @ Belerion · Robotics
Developing robust RL agents for robotic autonomy in complex, unpredictable environments. Bridging the gap from simulation to reality for dual-use applications.
PhD Researcher · Co-Founder @ Belerion · Robotics
Developing robust RL agents for robotic autonomy in complex, unpredictable environments. Bridging the gap from simulation to reality for dual-use applications.
PhD Researcher in Robotic Reinforcement Learning at ULiège & Co-founder of Belerion.
My research focuses on deploying RL agents from simulation to reality for autonomous systems that must operate in extreme conditions. I work on problems where decisions need to be fast and robust: high-speed robotic manipulation, autonomous drone navigation, and adversarial counter-UAS scenarios.
Supervisor
Prof. Damien Ernst, Montefiore Institute, University of Liège.
Collaborations
FN Herstal, John Cockerill Defence, and Thales.
Location
B28 Montefiore, Allée de la Découverte 10, Liège, Belgium.
Research Focus
NSPA — Capellen
Deploying RL agents from simulation to reality.
Investigating multi-agent reinforcement learning for autonomous pursuit-evasion. Training adversarial policies in high-fidelity simulation to study robust behaviors and sim-to-real transfer capabilities.
Researching robust tracking and control strategies for counter-UAS applications. Focusing on reinforcement learning algorithms that can adapt to dynamic targets and uncertain environments.
Optimizing robotic sorting efficiency through learned manipulation policies. Studying the application of reinforcement learning to improve pick-and-place performance in waste management contexts.
Peer-reviewed papers and preprints.
Abstract:As warfare becomes increasingly digital and autonomous, Reinforcement Learning (RL) has emerged as a promising technique for developing intelligent and adaptive drone behaviors. This paper identifies several remaining gaps in the current state of RL for drone warfare, focusing on bridging the gap between simulated training and real-world deployment.
Abstract:This paper presents a multi-agent reinforcement learning environment for drone combat built on IsaacLab. It includes an in-depth comparison between decentralized learning and self-play schemes in competitive settings, confirming the benefits of self-play for autonomous combat.
Abstract:This study proposes a novel approach based on reinforcement learning (RL) to enhance the sorting efficiency of scrap metal using delta robots and a Pick-and-Place (PaP) process. The approach uses a 3D-simulated environment to train an RL agent to perform throws of scrap metal into different bins, significantly improving sorting throughput.
Abstract:A study on exploiting reinforcement learning to teach an ABB Flexpicker robot to accurately throw objects into bins using a high-fidelity simulator.
Recent events and conferences.
Live demo of our advancements in autonomous drone interception system at the Paris Air Show.
Watch Demo
Participated in the Inno4Def drone hackathon organized by La Défense. Flying drones in simulation using Liftoff and developing autonomous strategies.
Watch DemoAcademic trajectory.
University of Liège — Montefiore Institute
Research on reinforcement learning for robotic systems with defence applications, supervised by Prof. Damien Ernst.
University of Liège
Specialization in machine learning and artificial intelligence. Thesis on reinforcement learning for robotic control.
University of Liège
Foundation in mathematics, physics, and computer science.
2024 — 2026
INFO0948 · University of Liège
Assisting in the Reinforcement Learning course, including tutorial sessions, student project supervision, and exam preparation. Covering value-based methods, policy gradients, actor-critic architectures, and multi-agent RL.
Beyond the lab.
Playing football with my team is how I disconnect.
Building and flying FPV drones allows me to experience flight from a different perspective.
08
Initiate communication sequence.
Belerion OS [Version 3.0.1]
(c) 2025 Arthur Louette. All rights reserved.
Accessing contact protocols...