A Robust MEMS-based GNSS/INS System for UAV Applications
Richard Deurloo, Valentin Barreau, Marnix Volckaert, Kristof Smolders, and Bei Huang, Septentrio Satellite Navigation, Belgium
In recent years commercial UAVs have moved from niche markets to a fast-growing professional market. UAVs are becoming the standard in large-scale surveying campaigns, serving as platforms for photogrammetry and LiDAR scanning.
For these applications a GNSS/INS system is a must, providing not only timing and position information, but also attitude. In the first place, the position and timing is needed for correct geo-referencing of the collected data. For photogrammetry, attitude can aid in stitching the photographs. For LiDAR scanning applications, high data-rate attitude information is needed to correct for platform motion and to remove artifacts from the point cloud.
The large amounts of data produced in these applications do not allow for manual corrections of the sensor data after the measurement campaign or after post-processing. This means that the GNSS/INS system needs to be robust and needs to provide reliable and accurate results, in real-time or post-processing. Robustness in the sense that the system needs to work every time and in all conditions, to limit downtime and to prevent having to re-do part of the measurement campaign. Reliable in the sense that the solution provided by the system needs to be trusted, to limit the need of manual inspection of the data. And accurate to limit degrading the final accuracy of the survey data.
In recent years, GNSS/INS systems based on MEMS IMUs have become available, fitting the strict weight and power budgets of UAV missions as well as driving down the costs. But how accuracy and reliable are they really? In the paper we will present Septentrio’s new GNSS/INS solution, specifically developed for UAV survey and inspection applications. The system will be benchmarked against most popular pre-existing commercial MEMS-based GNSS/INS systems.
All systems presented in this paper will use dual-antenna GNSS receivers. This adds GNSS attitude information to the INS filter, which can improve attitude accuracy and improves reliability at startup. But the accuracy of the GNSS attitude is highly dependent on the antenna baseline length and on the type of antenna (i.e. high precision vs. patch antenna). The impact of baseline length and antenna type will therefore be discussed first. For the benchmark all systems will share the same two antennas.
In collaboration with a survey UAV manufacturer the GNSS/INS systems will be benchmarked in multiple environments, typical for survey applications. The systems will be mounted on a large quad-copter UAV (12 kg) with a two-meter wide boom for the dual-antenna GNSS receivers. The UAV will be equipped with two reference systems: a high-end GNSS/INS system and a high-grade FARO LiDAR. The high-end GNSS/INS system will provide a direct reference for the position and attitude computed by the MEMS-based GNSS/INS systems under testing, allowing us to determine reliability and accuracy of these systems.
The LiDAR data collected with the UAV will be post-processed with the position and attitude solutions of each of the GNSS/INS systems under test to generate the LiDAR point cloud. This will allow us to compare of the point cloud solution for known features and repeated flight lines. In addition, terrestrial data will be collected with the same LiDAR mounted on the ground. This provides a millimeter-accurate geo-referenced model of known structures for absolute reference. Since the same LiDAR data is used for each of the GNSS/INS systems, this will serve as independent reference to determine the accuracy of each of the systems.