IDMC'07 20-21 Nov.

2007 Amir Kabir University

A Smoothing Algorithm based on Impulse Response for GPS Data Processing

M. R. Mosavi Assistant Professor, Department of Electrical Engineering, Behshahr University of Science and Technology, Behshahr 48518-78413, Iran Email: M_Mosavi@iust.ac.ir

Abstract An efficient spatial smoothing algorithm for filtering data while preserving spatial detail is obtained using the system function (impulse response) of the sensor. In contrast to the normal procedure for determining the filter coefficients for an arbitrary system functions, this technique does not involve the use of Fourier transforms. This paper uses from collected positioning and timing actual data for evaluating the performance of the algorithm. A simulator is designed and developed for this purpose. The spatial filtering coefficients are obtained analytically for the frequently applicable Gaussian system function. The efficiency of this algorithm is illustrated by filtering simulation, logs and spectral data. For real-time smoothing of data, a filter length of 4 times the vertical resolution is required. The experimental tests results on collected real data emphasize that positioning and timing RMS errors reduce to less than 0.4m and 65nsec, respectively. Keywords: GPS Data Processing, Smoothing, Impulse Response

is insufficient as a stand-alone positioning system. The accuracy which a GPS receiver can determine position and velocity of synchronize to GPS time is dependent on a number of factors. GPS error sources are allocated into three categories: the space segment, control segment, and user segment. The most dominant error source by far is Selective Availability (SA) which produces a one-sigma error of 32.3 meters. These error sources are for the civilian Coarse Acquisition (C/A) code. Real-time data smoothing must rely on computationally efficient (and therefore mathematically simple) methods to achieve good noise reduction without a serious loss of vertical detail. Adaptive filters were developed to recognize the logging data where signal level is substantially constant with depth and provide smoothing in these regions. There are also filters designed with information from other logs to assist in recognizing bed boundaries. Unfortunately, these techniques are still more computationally intensive than simple since filtering coefficients must be frequently recomputed. In this paper, an iterative smoothing algorithm is studied that is simple, computationally efficient and effective in preserving sharpness. Since this algorithm uses the system function (impulse response) of the sensor as a filter, the sharpness of the data is preserved to within the limits of this function while removing random noise. This filter can be applied to log data and spectral data (e.g. GPS data). This algorithm is a potential candidate for a real-time logging filter since it does not require more data than several times the vertical resolution of the sensor. Real log data is presented that studies the characteristics of this procedure. The results reveal that only a few iterations are required to obtain satisfactory results. Furthermore, the performance is not strongly dependent on the actual shape of the system function. Finally, a signal pass procedure is developed which is exactly equivalent to the iterative procedure. Spatial filtering coefficients are computed from the system function directly. In the case of a Gaussian system function, an analytic expression for the filter coefficients is derived.

1. Introduction
The Global Positioning System (GPS) is a satellitebased navigation system equipped with the receiver that provides the user with appropriate and accurate positioning-timing information anywhere on the globe. However, several errors are associated with the GPS measurement. It has superior long-term error performance, but poor short-term accuracy. For many applications, GPS

2. Experiments
Figure 1 shows a plan photograph of the internal board, which is supplied in a plastic case with connectors on a back plate. The used GPS receiver is a low cost and single frequency receiver manufactured by using Rockwell Company. The GPS receiver is also supplied with active antenna, cabling (both for antenna and PC link), mains adaptor, all software (including source code), and manuals. The GPS receiver to PC links adhere to

IDMC'07 20-21 Nov.2007 Amir Kabir University

the physical characteristics of the RS232 communications protocol with a default rate of 4800 bps (bits-per-second). The experimental test setup was implemented and installed in the Computer Control and Fuzzy Logic Research Lab in the Iran University of Science and Technology.

Table 2: 1000 positioning smoothing errors significance characteristics without SA

Parameters Max Min RMS Average Variance Standard Deviation X Component -0.1534 -0.0966 0.1223 -0.1216 0 0.0039 Y Component -0.5206 -0.1579 0.1881 -0.1876 0 0.0060 Z Component -0.2366 -0.1797 0.2112 -0.2110 0 0.0067

Table 3: 500 timing smoothing errors significance characteristics with SA

Parameters Max Min RMS Average Variance Standard Deviation Value [nsec] 114 0 45 18 4 2

Figure 1: Photograph of GPS receiver board and test setup The smoothing algorithm of section II was developed by paper author in Microsoft Visual Basic 6. It is used to smooth GPS receivers positioning and timing data. The software can analyze real data and can plot their results. The smoothing algorithm parameters selection is based on the experiments. The filtered noise is reduced by increasing the number of iterations. After the fourth iteration the sharpening effect starts to make the long range statistical correlations noticeable without visibly improving the bed boundary resolution. In order to evaluate the performance of the algorithm, Root Mean Square (RMS) was used as a measure of closeness between smoothed and observed values. Table 1 and Table 2 show smoothing errors (the difference between the smoothed and real values) statistical significance characteristics for 1000 positioning test data, before and after SA, respectively. Tables 3 and 4 present smoothing errors statistical significance characteristics for 500 timing test data, with and without SA, respectively.
Table 1: 1000 positioning smoothing errors significance characteristics with SA
Parameters Max Min RMS Average Variance Standard Deviation X Component -0.2722 -0.2153 0.2497 -0.2492 0.0001 0.0079 Y Component -0.4196 -0.0810 0.1058 -0.1049 0.0000 0.0033 Z Component -0.1406 -0.0837 0.1090 -0.1078 0.0000 0.0034

Table 4: 500 timing smoothing errors significance characteristics with SA

Parameters Max Min RMS Average Variance Standard Deviation Value [nsec] -56 -86 62 -62 9 3

Tables2, 3, 4, and 5 show that quality of GPS receiver data noise reduction is very good even under the strongly noise-corrupted conditions. Except that the proposed algorithm is more accurate, it is also less computation than many of the existing methods.

3. Conclusions
A simple filtering algorithm was investigated that results in optimal smoothing within the constraints of retaining good vertical detail after only 3-4 iterations. A refinement of this procedure involving higher order filter functions produced the equivalent operation in a single pass. In the special case of a Gaussian system function, an analytic expression for the spatial filtering coefficients was obtained. The filter is computationally efficient and is applicable to real-time smoothing of GPS receivers positioning and timing data. Quality of the obtained results is very good even under the strongly noise-corrupted conditions. It was shown that the algorithm reported in this paper presents a step toward the ideal, but unrealizable goal of prefect noise removal with prefect preservation of the underlying signal structure. The experimental test results on collected real data emphasize that positioning and timing RMS error reduce to less than 0.4m and 65nsec, respectively.