Florent Delgrange

Florent Delgrange is a postdoctoral researcher at Vrije Universiteit Brussel (VUB), working across multiple research units including R&D centraal and Federated labs AI and Robotics. He holds an unpaid teaching assignment in the School of Informatics and Applied Informatics.

His research focuses on the intersection of artificial intelligence, formal methods, and reinforcement learning. Key areas include:

• Formal verification of deep reinforcement learning policies
• Controller synthesis from learned policies
• Multi-agent systems with formal guarantees
• Model checking for AI systems
• Generative modeling in reinforcement learning frameworks

Dr. Delgrange's recent publications demonstrate a strong trend toward unifying reinforcement learning with formal methods to provide safety and reliability guarantees for AI agents. His work often combines deep learning approaches with formal verification techniques to ensure that learned policies satisfy specific temporal logic constraints.

Dr. Delgrange is actively involved in the research community:

• Peer reviewer for AAAI Conference on Artificial Intelligence (2023)
• Member of programme committee for European Workshop on Reinforcement Learning (2023)
• Peer reviewer for International Conference on Autonomous Agents and Multiagent Systems (2023)
• Participant in multiple reinforcement learning workshops (2022-2023)

He currently leads the research project "Reliable and Adaptive Learning Agents: Unifying Reinforcement Learning and Reactive Synthesis" (OZR4417) from 2025-2026, which aims to bridge the gap between reinforcement learning and formal synthesis techniques.