Previous Abstract Return to Session D3 Next Abstract

Session D3: Aerial Vehicle Navigation

Small UAV’s Attitude Estimation with Tightly Coupled Low-cost GNSS/INS Integration using Multiple GNSS Receivers
Marton Farkas, Hungarian Academy of Sciences, Budapest University of Technology and Economics, Hungary; Balint Vanek, Institute for Computer Science and Control, Hungarian Academy of Sciences, Hungary; Szabolcs Rozsa, Faculty of Civil Engineering, Budapest University of Technology and Economics, Turkey
Location: Spyglass

The spread of the unmanned aerial vehicles (UAVs) is indisputable. The most important commercial UAV application fields are the photogrammetry, surveying, mapping, inspections, monitoring, surveillance, precision agricultural, delivery. It is necessary to improve the navigation methods to keep the safety of the presumably crowded airspace in the future, which is inevitable with the continuously growing UAV segment. The attitude angle estimation has a prominent role in this improvement, since.it is one of the most important information of the flight. It is used for the automatized control of these vehicles, thus it must be estimated redundantly.
This paper presents a tightly coupled multiple GNSS/INS sensor integration methodology for attitude estimation of small UAVs. The attitude angles are determined from fusion of single frequency, single baseline multi-constellation (GPS, Glonass, Galileo) GNSS code and phase measurements, accelerometer, gyroscope and magnetometer data using an Extended Kalman Filter. The GNSS attitude estimation part of the algorithm is based on the double-differenced code and phase measurements using the known baseline length between the antennas as a constraint. The two different attitude solution could complement each other. The inertial sensors provide the measurements with high sampling rate, but the bias errors of the gyroscope and the accelerometer sensors reduce the reliability of this solution, especially in case of low-cost sensors. The bias errors can be corrected for long term with the application of multiple GNSS receivers. The inertial sensors give information about the attitude angles in bad GNSS reception scenarios for example in case of highly dynamic maneuvers. The key of the GNSS attitude estimation is the integer ambiguity resolution. The fusionated attitude estimation algorithm also provides the float solution of the double-differenced ambiguities, which are the inputs of the constrained least-squares ambiguity decorrelation adjustment method. This non-linear optimization uses the additional information that the baseline length between the antennas is known. The search space bounding decreases the computational load to achieve real time attitude solutions. The successful integer ambiguity resolution is a key factor to achieve the required accuracy of the attitude estimation.
This methodology is used for the estimation of the attitude angles of a small UAV. The low-cost single frequency multi-constellation (GPS, Glonass, Galileo) GNSS receivers and the inertial sensors were placed in the UAV’s nacelles and the fuselage, antennas were attached at the wings. The real flight data is used in a tightly coupled attitude determination algorithm in post-processing mode. This fused attitude solution is compared to the simple multi-constellation GNSS and IMU attitude solutions. It is examined how the integer ambiguity resolution ratio changes depending on which estimation is used (tightly coupled multiple GNSS/INS or only multiple GNSS). It is also examined how the different GNSS constellations influence the solution reliability and the integer ambiguity resolution success rate.
The first results show, that the multi-GNSS attitude determination algorithm provided the attitude angles with the accuracy of 1-5 degrees compared to the independent IMU solution, that means that the accuracy of the GNSS attitude angles are comparable to the accuracy of the attitude angles obtained from the IMU. Moreover it was also confirmed, that the availability of GNSS attitude estimation strongly depends on the flight dynamics, which is usually much higher in case of UAVs compared to commercial aircrafts.



Previous Abstract Return to Session D3 Next Abstract