GNSS-INS
An overview of GNSS + Inertial navigation and its applications in UAV-based remote sensing
What is GNSS?
Global Navigation Satellite System (GNSS) refers to a system of spaceborne satellites that provide signals transmitting geolocation and timing data to GNSS receivers. The receivers then use this data to determine their real-time 3D location and date/time at a relatively low frequency (1-10Hz) and meter-scale accuracy. Raw GNSS observations from a roving GNSS receiver can be stored and post-processed with updated satellite geolocation (ephemeris), atmospheric corrections, and fixed base station reference data to produce a more accurate trajectory estimate than possible in real-time.
GNSS systems provide global coverage by definition, but component satellite constellations are operated by different entities. The most well known GNSS, Global Positioning System (GPS), is operated by the USA. Other constellations are listed below.
GNSS Constellations and Operating Countries
Global Positioning System (GPS)
US Government
Global Navigation Satellite System (GLONASS)
Russian Government
Galileo
EU/GSA
BeiDou
Chinese Government
What is an INS?
An Inertial Navigation System (INS) integrates measurements from an Inertial Measurement Unit (IMU) over time to estimate position, velocity, and orientation when given an initial starting condition. Because IMU data is available at a relatively high frequency, Inertial Navigation Systems estimate position and orientation at a much higher rate than GNSS, around 200Hz.
The main components of an IMU- gyroscopes and accelerometers, determine the change in orientation and velocity of an object. Some IMUs also employ a magnetometer to measure orientation relative to the Earth's magnetic field.
GNSS-INS
In order to achieve the best position and orientation estimate, a combined GNSS-Inertial Navigation System (GNSS-INS) is used. The fusion of GNSS and IMU data is essential to create an advanced GNSS + Inertial Navigation System that offers high precision, accuracy, and reliability. This fused system leverages the strengths of both GNSS and IMU technologies to overcome their respective weaknesses, providing more accurate position and orientation in real-time, and survey-grade accuracy when post-processed.
Benefits of GNSS-Inertial Navigation
Continuous Positioning: INS can maintain positioning when GNSS signal is momentarily lost.
High Update Rate: The high-frequency data from the IMU complements the slower GNSS update rate, ensuring smooth and continuous motion tracking.
Reduced Drift: GNSS helps to correct the long-term drift that is inherent in the IMU data.
Better Accuracy: Combining both GNSS and IMU data can provide centimeter-level positioning when post-processed.
GNSS-INS for Direct Georeferencing
The GNSS-INS is the key component allowing orthorectification and reconstruction of image and LiDAR data on GRYFN systems. In a process called Direct Georeferencing (DG), sensor exterior orientation parameters (EOPs) are measured directly by the GNSS-INS and are used to project data from an initial sensor reference frame into a common "real world" mapping frame. These EOPs describe a sensor's position () and orientation () with respect to Earth.
Line-scan hyperspectral and LiDAR sensors rely purely on DG to process raw data and produce orthorectified mosaics and point clouds. Frame camera data processed in GRYFN's software, however, follows a more traditional photogrammetric approach, but is augmented with GNSS-INS position and orientation to remove the need for Ground Control Points (GCPs) and improve weak image feature matching often experienced when processing data over a homogenous scene, such as large agricultural fields.
Orthorectification and Boresight
Because the GNSS-INS is not co-located and perfectly aligned with the reference coordinate system of each sensor, translation and rotation offsets must be applied to correctly position pixel and point data on Earth's surface or in 3D space. The process of determining the geometric relationship between the GNSS-INS and each sensor is called Boresight Calibration and is discussed in more detail in the Boresight Calibrationsection of this wiki.
Alignment/Initialization
GNSS-INS systems require an alignment or initialization phase to establish baselines for position/velocity, attitude, and sensor bias estimates that the Extended Kalman Filter (EKF) relies on for accurate navigation solutions. During this phase, the system uses static and/or dynamic maneuvers to observe the relationship between the IMU’s inertial measurements and the GNSS-derived positions and velocities. These observations allow the EKF to resolve sensor biases, including accelerometer and gyro offsets, and to determine the initial orientation of the IMU relative to the Earth. Without a properly executed alignment, the EKF cannot correctly integrate the high-rate inertial measurements with GNSS data, leading to accumulated drift, errors in velocity and position, and degraded attitude estimation.
While best practices vary from one manufacturer to another, and users should inform themselves on best practices with their specific system, the key to the alignment phase is that dynamics are performed. For some systems, this requires accelerated turns, while others prefer high linear acceleration. For most systems, regardless of procedure, 5 to 10 seconds of dynamic movement is sufficient. With slower accelerating aircraft, users should try to maximize dynamic alignment time.
All GNSS-INS solutions GRYFN integrates post-process trajectories both forwards and backwards. Dynamic alignments must be performed at the beginning and end of each flight so the post-processing kinematics have alignments procedures at the beginning of the trajectory for both the forward and backwards processing.
PPK vs RTK
Post-Processing Kinematics (PPK) and Real Time Kinematics (RTK) are two different correction methods commonly used in remote sensing to improve the positional accuracy of reconstructed data products. Both approaches leverage GNSS reference station(s) to correct the inherent error in raw GNSS observations. We can see in the table below that real time GNSS-INS data, without PPK or RTK correction, can yield positional accuracies on the meter scale, while processed trajectories can achieve centimeter level precision. We'll expand on precision and accuracy of each methodology in subsequent paragraphs.
With RTK corrections, a link must be established between the GNSS receiver in the payload and a ground reference station(s). Therefore, base and satellite corrections of GNSS observations are transmitted to the payload in real time. This requires: 1) GNSS receiver is capable of receiving RTK corrections, 2) Users link their GNSS receiver to a reference station(s) (either local CORS station(s) or user-setup base station). To connect to CORS stations, users must be in close proximity to a base station (< 10-20km) or in relatively close proximity to a network of base stations (< 50-70km). In remote locations without access to a single reference station or network of reference stations, users must setup their own base and record data for a considerable time before the mission begins. While this process can allow for accurate and faster turnaround times with DG processing, it can require considerable setup time every mission, and may not be as accurate as PPK processing.
PPK instead applies corrections during the post-processing phase. PPK can take advantage of longer duration base data and precise satellite ephemeris data that typically allows for more accurate trajectories than RTK. PPK also limits crucial time in the field with user involvement, as no real-time link to reference stations is required. With some GNSS-INS manufacturers, PPK can also eliminate the need for a user-setup base station if CORS stations are not available by leveraging proprietary satellite-based correction methods, such as Trimble Applanix' PP-RTX. The downside to the PPK approach is data turnaround time, as users will be unable to achieve the same level of accuracy as RTK if attempting to process at the field-edge.
RTK provides immediate access to corrections which can allow for faster confirmation and analysis of data collection, but requires user-intervention in setup for every mission, and will not provide nearly as accurate trajectory solutions as PPK. PPK will provide the most accurate trajectory solutions, provides a consistent, repeatable workflow with no user-intervention in the field, but requires an additional step in post-processing (which is automated through GRYFN Processing Tool). GRYFN relies on PPK methods as it streamlines workflows and provides the highest level of accuracy and precision, but are evaluating RTK integrations for relatively accurate edge computing capabilities.
FAQ
How accurate are the GNSS-INS solutions in GRYFN systems?
GNSS-INS solution accuracy depends heavily on positioning mode and other factors. Typical performance of real-time and post-processed solutions is given below for a UAV in flight conditions with good GNSS reception.
Typical GNSS-INS Performance
Position (m)
1.2 - 3.0
0.01 - 0.05
Roll & Pitch (deg)
0.03 - 0.04
0.015 - 0.025
Heading (deg)
0.1 - 0.3
0.035 - 0.080
Why don't GRYFN systems use the GNSS receiver or IMU on my UAV?
GRYFN chooses to use independent GNSS-INS hardware to ensure the highest quality data is available for post-processing. By using a survey-grade GNSS-INS and post-processing software, much higher accuracy and repeatability can be achieved, resulting in the best possible data products and alignment.
What can cause issues with the GNSS-INS?
Here are some things to do when using a GNSS-INS:
Ensure the GNSS antenna(s) have a clear view of the sky
Both from overhead obstructions, as well as from the aircraft body during attitude changes
Avoid long periods of static data (longer than 2-3 minutes after IMU initialization)
Ensure the GNSS-INS is aligned and initialized before starting data collection
Avoid Electromagnetic Interference (EMI) or unnatural magnetic fields that may interfere with the GPS or IMU
Last updated