Kilometer-Scale GNSS-Denied UAV Navigation via Heightmap Gradients: A Winning System from the SPRIN-D Challenge

TL;DR

This work tackles long-range GNSS-denied UAV navigation by fusing lightweight, geo-data–driven drift correction with odometry. The authors introduce a heightmap-gradient matching approach that aligns LiDAR-derived local heightmaps to a prior geodata heightmap and fuse this similarity with odometry in a clustered particle filter, all running onboard on CPU-only hardware. The system demonstrates kilometer-scale, fully autonomous flights across urban, forest, and open-field environments, achieving an RMSE below 11 m and placing first in the SPRIN-D challenge. The study emphasizes that robust re-localization and structured diagnostics under uncertainty are more critical than maintaining ultra-low instantaneous drift, offering a practical blueprint for GNSS-denied UAV autonomy in real-world deployments.

Abstract

Reliable long-range flight of unmanned aerial vehicles (UAVs) in GNSS-denied environments is challenging: integrating odometry leads to drift, loop closures are unavailable in previously unseen areas and embedded platforms provide limited computational power. We present a fully onboard UAV system developed for the SPRIN-D Funke Fully Autonomous Flight Challenge, which required 9 km long-range waypoint navigation below 25 m AGL (Above Ground Level) without GNSS or prior dense mapping. The system integrates perception, mapping, planning, and control with a lightweight drift-correction method that matches LiDAR-derived local heightmaps to a prior geo-data heightmap via gradient-template matching and fuses the evidence with odometry in a clustered particle filter. Deployed during the competition, the system executed kilometer-scale flights across urban, forest, and open-field terrain and reduced drift substantially relative to raw odometry, while running in real time on CPU-only hardware. We describe the system architecture, the localization pipeline, and the competition evaluation, and we report practical insights from field deployment that inform the design of GNSS-denied UAV autonomy.

Figures (12)

• Figure 1: The autonomous system for long-range GNSS-denied flight presented in this paper achieved 1st place in the international Fully Autonomous Flight Challenge.
• Figure 2: System overview. Sensor inputs from LiDAR, cameras, IMU, and compass feed modules for visual–inertial odometry, obstacle mapping, and flag detection. A similarity-based localization module fuses odometry with prior geodata in a particle filter. Mission control coordinates planning and control for waypoint-based flight.
• Figure 3: Overview of the UAV platform and integrated sensor suite. The multirotor airframe carries a Livox Mid-360 LiDAR for mapping and localization, an Intel RealSense D435 depth camera for short-range obstacle perception, a Bluefox RGB camera and IMU for VIO, and an onboard Intel NUC computer for real-time processing.
• Figure 4: Custom 3D-printed mount with integrated sensors, mechanically decoupled by silent blocks (blue) to dampen vibrations and improve VIO robustness.
• Figure 5: The mission control state machine