AIRSHIP & FALCON-S

Overview

AIRSHIP is a Horizon Europe project developing autonomous systems for Unmanned Wing-in-Ground (WIG) Vehicles — a class of fixed-wing aircraft that fly close to the surface, exploiting ground-effect aerodynamics for fuel efficiency and payload capacity in maritime and coastal applications.

As Lead Research Scientist on RL-based autonomous navigation, I lead the development of advanced Guidance, Navigation, and Control (GNC) strategies that combine optimal control and reinforcement learning policies for autonomous low-altitude flight.

Out of this work, I built FALCON-S — a high-fidelity simulation and learning framework for fixed-wing robots in ground effect, currently under review at RSS 2026. FALCON-S is the technical backbone behind AIRSHIP's autonomous GNC pipeline.

FALCON-S — Simulation & Learning Framework

Fixed-Wing Robots in Ground Effect: A High-Fidelity Aerodynamic Simulation Framework for Flight Control

FALCON-S addresses the unique aerodynamic challenges of near-surface flight, where ground-effect interactions create highly nonlinear and rapidly varying dynamics that classical controllers struggle to handle robustly. The framework provides a unified environment for training, evaluating, and benchmarking autonomous controllers across realistic WIG-vehicle flight regimes.

🛩️ Ground-Effect Aerodynamics

Modular flight-dynamics simulator (MATLAB & Python) capturing the highly nonlinear ground-effect aerodynamics of fixed-wing vehicles flying close to the surface.

🤖 Deep RL Controllers

Deep reinforcement learning policies for autonomous low-altitude flight, robust to ground-proximity aerodynamic perturbations, sensor noise, and wind disturbances.

📊 Classical GNC Benchmarks

Systematic benchmarking of RL policies against classical baselines — PID, LQR, and MPPI — across the full envelope of near-surface flight regimes.

Publication: Fixed-Wing Robots in Ground Effect: A High-Fidelity Aerodynamic Simulation Framework for Flight Control — Matteo El-Hariry, Pedro Lima, Andrej Orsula, Matthieu Geist, Miguel Olivares-Mendez. RSS 2026 (under review)

Research Contributions

Simulation Framework

Modular flight dynamics: Novel simulator operating in both MATLAB and Python, capturing ground-effect aerodynamics with parametrizable airframe configurations.
Realistic disturbance modeling: Wind disturbances, sensor noise, and surface proximity effects to enable training under realistic maritime conditions.

Autonomous GNC Policies

Deep RL controllers: Designed and trained policies for autonomous low-altitude flight in ground effect, targeting robust maritime and coastal operation.
Classical baseline comparisons: Rigorous benchmarking against PID, LQR, and MPPI controllers to evaluate learning-based advantages.
End goal: Enable fully autonomous WIG-vehicle operation — a class of aircraft with no existing autonomous GNC solution.

Resources

AIRSHIP Project Website Project Objectives

Autonomous GNC and High-Fidelity Simulation for Unmanned Wing-in-Ground Vehicles