Airborne Gravimetry by GPS/INS A Comparison of Filtering Methods

Youcef Hammada

Abstract:	During the last five years, research in airborne gravimetry has been very active due to the fact that aircraft acceleration can now be determined with high accuracy from GPS carrier phase and phase rate measurements. Airborne gravimetry determines the gravity disturbance by differencing a time series of gyro-stabilized accelerometer measurements from an inertial navigation system (INS) and a time series of GPS derived accelerations. The difference is the sum of the gravity disturbance and system noise. The airborne implementation of this principle is so attractive because it allows to measure the gravity signal in large areas in relatively short time. Both measurements are affected by large noise with different frequency components. Consequently,the gravity disturbance signal is buried in considerable noise and its frequency spectrum resolution will depend on the signal to noise ratio in different frequency bands. The challenge is to derive the most appropriate filtering scheme to recover the useful signal from the noisy measurement. Three approaches are considered and compared in this paper. Finite impulse response low-pass filtering which is basically a frequency domain technique. Kalman filtering with a shaping filter as a stochastic model for the gravity disturbance. Deterministic model filtering that approximates the gravity disturbance by a simple function, such as a constant or a ramp, for a short time interval. The latter approach divides the complete estimation cycle into a series of estimation cycles of short duration, and is therefore quite different from the other two approaches which consider the whole data span as one function. The same set of airborne data was used to estimate gravity disturbances by each of the three methods. The gravity disturbance estimates were compared to values obtained through upward continuation of terrain gravity disturbances, used as reference. The deterministic model filtering method showed the expected discontinuities between estimation cycles. After smoothing by an interpolating polynomial, the gravity disturbance estimates reproduced the reference trend. The frequency domain method performance depends on the sensor errors at different frequencies. Again, the gravity disturbance estimates reproduced the reference trend. The Kalman filter method,based on a stochastic model for the gravity disturbance, yielded the best results after a backward optimal Kalman filter smoothing step. The results are preliminary and some possible limitations of the analysis are discussed in the last section.
Published in:	Proceedings of the 8th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GPS 1995) September 12 - 15, 1995 Palm Springs, CA
Pages:	339 - 347
Cite this article:	Hammada, Youcef, "Airborne Gravimetry by GPS/INS A Comparison of Filtering Methods," Proceedings of the 8th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GPS 1995), Palm Springs, CA, September 1995, pp. 339-347.
Full Paper:	ION Members/Non-Members: 1 Download Credit Sign In