Jason Hamilton
Product Manager
Email: [email protected]
Rod MacLeod
South Asia and Australia Regional Sales Manager, NovAtel Inc
Email: [email protected]
More and more of our physical world is being geo-referenced and the data used for a wide variety of applications by companies, governments and militaries; or offered to the public through sites like Google Maps and Microsoft Virtual Earth. Spatial data has become so readily available that it is now assumed precise and accurate, but how is that accuracy achieved?
Until recently, only 2D representations of the Earth were widely available. Airborne survey systems utilised GNSS technology for geo-referencing airborne imagery to create geo-referenced digital mosaics. Airborne flight conditions are ideal for GNSS systems as the satellite signals are nearly always available and have high reliability. Precision is improved if the data is post-processed with a local base station or with precise point positioning using precise clock and orbit data.
In the last few years, traditional overhead imagery has begun to be augmented with data collected at street-level. Imagery from the ground gives a more intuitive view of the surroundings and opens many new uses for visualisation of surroundings. Extraction of features from groundbased imagery also offers an efficient method of collecting GIS data or monitoring environmental changes. However, ground-based data collection for both imagery and LiDAR places increased demands on positioning systems. Unlike airborne data collection, ground based systems operate through frequent partial or total blockages of the GNSS satellite signals. Obstructions like buildings and heavy tree cover limit the availability of GNSS signals and create multipath that can degrade the accuracy of the position solution. Even with post-processing techniques, GNSS alone does not offer enough availability or reliability for ground mapping applications.
One solution to this problem is to aid the GNSS system with a complementary technology to improve reliability and accuracy. A good GNSS augmentation choice is an Inertial Navigation System (INS), which uses measurements from an inertial measurement unit (IMU) to compute a position, velocity and attitude solu- tion. Unlike GNSS, INS requires no external inputs, so the accuracy of the system does not vary with the environment.
GNSS positioning and inertial navigation are complementary technologies. GNSS receivers track satellite signals to compute a position that can be as accurate as 1cm, but do not provide any attitude (roll, pitch or heading) information for the vehicle. GNSS accuracy is also subject to the satellites in view and can be degraded or unavailable if satellites are blocked. An INS integrates measurements from gyroscopes and accelerometers to compute attitude, velocity and very stable relative position.
However, INS solutions drift over time due to accumulation of sensor measurement errors. In general, the higher the grade of the IMU, the lower is the drift rate. When combined together, GNSS is used to calibrate the errors in the INS and INS is used to compensate for periods when GNSS is unreliable. The GNSS/INS combination is also able to provide high rate accurate roll, pitch and heading.
An example of a commercial, off-theshelf GNSS/INS combination is NovAtel’s SPAN™ (Synchronised Position, Velocity and Attitude) product line, which is unique because of the way GNSS and IMU data is combined. SPAN tightly integrates the data inside the GNSS receiver, providing several advantages. Unlike loosely-coupled GNSS/INS systems that integrate INS and GNSS in the solution domain, SPAN integrates the data together in the measurement domain to create a very tightly coupled system. The stability of the blended GNSS/INS solution is also used to aid the GNSS functionality of the receiver. GNSS algorithms like the RTK and Lband correction filters are aided with the GNSS/INS solution to achieve faster solution convergence. Furthermore, the GNSS/INS solution is used to aid the signal tracking loops, improving signal reacquisition during difficult GNSS conditions. The result is nearly instantaneous GNSS signal reacquisition and more accurate satellite observations.
The benefits of a tightly coupled system can be measured in two significant ways; the satellite signal reacquisition time and the drift of the inertial solution during partial GNSS outages. Signal tracking aiding allows for faster signal reacquisition after GNSS outages. Figure 2 shows the cumulative histogram of GNSS signal reacquisition with and without aiding from the SPAN solution. Faster signal reacquisition results in more GNSS observations to use to constrain any inertial errors. Figure 3 shows the number of satellites tracked during an urban mapping data collection from a loosely coupled Vs a tightly coupled system. Because GNSS and IMU data is combined at the measurement level in SPAN, the GNSS data is used even when fewer than 4 satellites are available.
In a loosely coupled system, no update information would be used when the tracked satellite count dropped below 4, as no GNSS solution would be possible. Figure 5 shows the benefit of using the carrier phase data as a constraint to the inertial solution drift, when fewer than 4 satellites are available. The red 0 Phase line shows the drift of the solution during a complete GNSS outage. This drift is dominated by the accumulating errors from IMU measurements. If two satellites are available, SPAN will use the data to constrain the drift as shown in the 1 Phase green line. If another satellite is available, the blue 2 Phase line shows even more significant improvement.
SPAN is a real-time system that operates in multiple modes, depending on the required accuracy of the application. In stand-alone mode with no GNSS correction data applied, position accuracy of <1.5m can be achieved. If local base station corrections or satellite-based augmentation systems like Omnistar HP are input into the system, cm-level positioning can be achieved. An optional wheel sensor input is also available as an additional constraint during GNSS outages.
Many mapping applications do not require a real-time solution for the generation of their final product. These applications can benefit from post-processing, where the GNSS and IMU measurement data is combined postmission. Post-processing has many practical and performance advantages. Data can be processed forwards and backwards in time, thus minimising the drift during GNSS outages. A solution smoother further reduces solution error. Absolute centimetre-level accuracy can be achieved using data from a local or virtual reference station (VRS). Figure 6 shows the improvement on position error growth during GNSS outages by processing forwards and backwards in time.
NovAtel’s Inertial Explorer® (IE) software package offers easy import and processing of SPAN measurement data. The performance improvement from post-processing is most evident in the position drift during a full 60s outage. For a SPAN system operating with an iMAR FSAS IMU, the solution drift over 60s in real-time and post-processing is shown in Table 1.
All of these features result in an improvement to the solution accuracy and integrity. The difference between a GNSS-only system and a SPAN solution is clearly noticeable in an urban canyon test. Figure 7 shows a trajectory view of a GPS-only solution and of a SPAN GPS/INS solution. While the GNSS solution is often in error because of poor satellite geometry and multipath signals, the SPAN GPS/INS solution offers continuously precise positioning.
Conclusion
GNSS-only systems do not provide suitable positioning accuracies and availability for the new applications in ground mapping where multi-sensor observations are made at high data rates in highly obstructed areas. A GNSS/INS combination provides a solution to the problem and SPAN, with its tightly coupled architecture in combination with Inertial Explorer post-processing software, offers an “off the shelf” system with maximum performance to the user.
