Stable Multi-Drone GNSS Tracking System for Marine Robots
Shuo Wen, Edwin Meriaux, Mariana Sosa Guzmán, Zhizun Wang, Junming(Clark) Shi, Gregory Dudek
AI summary
Problem
GNSS signals vanish underwater, forcing marine robots to rely on drift-prone inertial sensors or infrastructure-heavy acoustic systems. Single-drone offboard tracking further suffers from occlusions and frequent tracking loss.
Approach
The system fuses lightweight visual detection, geometric triangulation, and a confidence-weighted Extended Kalman Filter across multiple drones, while a cross-drone ID alignment algorithm ensures consistent robot identification across overlapping views.
Key results
- Sub-2m mean localization error across diverse tracking scenarios
- Accuracy improves with aerial coverage (0.94 m with three drones vs. 1.11 m with one)
- Robust tracking through rapid direction changes via hybrid IOU-GNSS matching
- Open-source, scalable framework for low-cost offboard marine robot positioning
Why it matters
Provides a reliable, infrastructure-free positioning solution for marine robotics, aquaculture monitoring, and search-and-rescue operations where traditional localization fails.
Abstract
Stable and accurate tracking is essential for ma- rine robotics, yet Global Navigation Satellite System (GNSS) signals vanish immediately below the sea surface. Traditional alternatives suffer from error accumulation, high computational demands, or infrastructure dependence. In this work, we present a multi-drone GNSS-based tracking system for sur- face and near-surface marine robots. Our approach combines efficient visual detection, lightweight multi-object tracking, GNSS-based triangulation, and a confidence-weighted Extended Kalman Filter (EKF) to provide stable GNSS estimation in real time. We further introduce a cross-drone tracking ID alignment algorithm that enforces global consistency across views, enabling robust multi-robot tracking with cooperative aerial coverage. We validate our system in diversified complex settings to show the accuracy and robustness of the proposed algorithm.