Multi-sensor data fusion for state estimation of unmanned surface vehicles

Abstract

Small-scale Unmanned Surface Vehicles (USVs) offer cost-effective platforms for Maritime Autonomous Surface Ships research, enabling deployment early in the design process. However, their affordability comes with a trade-off in equipment quality, imposing challenges for navigational safety. This study aims to customize a state estimation method for USV navigation by integrating an Extended Kalman Filter (EKF) with data fusion from a Global Navigation Satellite System module and an Inertial Measurement Unit. Method Verification is performed using mock data from an established automotive benchmark, demonstrates improved accuracy relative to over raw reference data. To further evaluate the approach, a compact and cost-effective data acquisition setup is designed and deployed on a USV operating in calm inland water conditions. The acquired real-world data are employed to validate the method, demonstrating its effectiveness in smoothening the estimated trajectory and mitigating discrete jumps between subsequent position measurements. Results also highlight Challenges, including reduced accuracy during initialization and under aggressive vehicle movement. This study provides insights into implementing effective state estimation methods for USVs, hence contributing to advancing their navigation.