|
Establishment & Testing of Dubai Virtual Reference System (DVRS) National GPS-RTK Network 4. Assessment of DVRS The DVRS system is fully tested with regard to its accuracy, speed of operation and optimum utilization of advance technology. It is also successfully implemented in day to day survey activities of DM survey section. It is being used for the various activities in the Department such as demarcation of plots and alignments, providing survey controls, GPS levelling and DTM generation. The testing of DVR system was carried out at various stages and for the various factors. The system was tested for its positional accuracy and network coverage. A random check for the telephone connection time, correction receiving time and rover initialization time were also being carried out. 4.1 Field Testing of The DVRS Testing the DVRS network in this study included investigating its availability, absolute and relative accuracy, precision and compatibility of results, reliability, and robustness. In addition, the performance of the network approach against the traditional approach of employing a single reference station was investigated. 4.1.1 Investigating Positioning Accuracy Using the DVRS Network The DVRS availability and general positioning accuracy were first examined at three independent locations, where coordinates of ten points at each location were determined. The first test was performed in an urban area, located a few kilometers away from the (Al Qusies) station. The second test was close to the (Al Lusali) station near to the centre of the network, and the third was on the southern border of the network coverage area, which is approximately 11.53 Km from the nearest reference station. The estimated average standard deviation of point coordinates for the three tests are given in Table 1, which were less than 2 and 3 cm for the 2D and height, respectively. 
To test the network absolute and relative accuracy, a set of 13 with distances ranging between 18.875m and 36.968m were established approximately 9.2 km away from the 'Al Qusies' control station. The layout of the test points is illustrated in Figure 4.1 Their coordinates have first been determined using a total station to an accuracy of less than 1 cm based on a least-squares adjustment approach integrating the angular and distance measurements. After that, the point coordinates were estimated in real time using the DVRS data. Three independent surveys were carried out with time spacing of 1-2 hrs to allow for significant changes in satellite geometry and atmospheric conditions. The PDOP values ranged from 1.9 to 3.92. The absolute (external) accuracy of the DVRS coordinates estimation is expressed as the differences between results of the DVRS survey and the total station. The differences are estimated taking the total station results as the reference. The 3D (spatial) differences ranged between 0.81cm and 3.61 cm. The estimated differences at each point for the three tests are illustrated in Figure 4.2  Figure 4.1 Layout of the test points  Figure 4.2 3D external accuracy of the DVRS The accuracy of relative positioning was evaluated by studying differences between distances derived from the DVRS estimated point coordinates against their precise values, which were independently determined using a calibrated stainless steel measuring tape. Figure 4.3 depicts the values of distance discrepancies. Table 4.2 shows their average and maximum values as well as their standard deviations. As can be seen, the differences were within 1 cm for the three tests, and the maximum discrepancy found was 2.2 cm. The standard deviations ranged from 0.4 cm to 2.2 cm.  Figure 4.3 Differences in length computation 
4.1.2 Testing of Precision and Compatibility of DVRS Results To test the achievable positioning precision from the DVRS data, the results of the three tests were compared, recalling that they are independent as they were carried out in different times. This comparison would not only test the compatibility of DVRS results, but also indicate system availability at different observing sessions (times). To examine the compatibility of results, discrepancies in coordinate estimation for the 13 test points were computed between the three tests, taking the first one as the reference. Figures 4.4 and 4.5 shows the discrepancies in coordinate estimation between the second and first tests, and between the third and first tests, respectively. The planemetric ( E & N) discrepancies were generally less than 5 cm, while the height differences were generally less than 6 cm, although it reached at some instances 10 cm. The average and maximum discrepancy values are given in Table 4.3 The table also gives the coordinate standard deviations (s) as estimated from the discrepancies  Figure 4.4 Discrepancies between the 2nd and 1st tests  Figure 4.5 Discrepancies between the 3rd and 1st tests 
To statistically examine the compatibility and the consistency of internal precision between the three tests, the discrepancies of the results of the second and third tests from the first test were tested using the Fisher (F) test. The test can be directly applied by testing the significance of the difference between any two variance estimates related to two samples of coordinates discrepancies. The ratio of the two sample variances should lie within a specified confidence region, such that ref Wübbena. G., , June 5-8, 2001: 
Where, s21 is the estimated variance of the sample that has the larger standard deviation (s), and s22 is the estimated variance of the sample that has the smaller value of s. The confidence boundaries can be evaluated from the tabulated values of Fisher distribution with (µ) probability that can be taken as 95% and (df1 and df2) degrees of freedom corresponding to the first and second samples of position discrepancies, respectively. This gives 2.69 and 0.11 for the upper and lower limits for the test in hand. The spatial (3 D) precision of the three tests, expressed as the average values of the 3D standard deviations of the computed 13 points from the 3 independent surveys, are 0.0264m, 0.0273m, and 0.0282m, respectively. Consequently, the examined ratios for the second and first tests, and for the third and first tests are 1.07 and 1.14, respectively, which lie within the upper and lower boundaries of the compatibility test, thus, passing the test. This means that the three tests are consistent in terms of their internal precision of the final output results. This can also be statistically interpreted that the three tests are statistically compatible (equivalent). 4.1.3 Investigating System Reliability and Robustness In order to test the DVRS output reliability and robustness, particularly in case of failure of one of the reference stations, a set of ten points were surveyed 2 km away from the 'Al Lusayli' reference station using the DVRS RTK data under two scenarios. In the first, the data of all five reference stations including " Al Lusayli" were incorporated in the computation of the phase measurements corrections. In the second case, the measurements of the " Al Lusayli" reference station were eliminated in the process of computing the DVRS data, resembling a case of failure of this station. In this case, the nearest control station to the survey area was approximately 25 km away. Usually, in the classical single reference RTK approach, such a case can only be solved using a float ambiguity resolution, leading mostly to a positioning accuracy at the decimeter level. Point coordinates in the case of eliminating the 'Al Lusayli' were also processed in two sessions. The first had a low PDOP values (less than 5), and the second had higher PDOP values. The second case represents one of the worst scenarios that can occur while surveying with the DVRS network. Table 4.4 gives the average and maximum coordinates standard deviations in all scenarios. All DVRS stations low PDOP LSLY is disabledlow PDOP LSLY is disabledHigh PDOP 
The results of Table 4.4 show the importance of employing the network concept. In the case of incorporating all reference stations, the standard deviations of the estimated point coordinates were generally less than 2 cm. In the second case of 'failure' of the LSLY station and under low PDOP values, the standard deviations were slightly larger, but stay at the cm level, where they did not exceed 3 cm for 2D positioning and 5 cm for height determination. Even in the case of high PDOP values, the standard deviations did not exceed 5 cm and 9 cm for 2D and height estimation, respectively. This proves that the DVRS performance is more reliable compared to the classical approach of employing a single reference station. In addition, the system proves to be robust in case of failure of one of its stations. 4.1.4 Comparison with Results of A Single Reference Station for Short Ranges The achievable positioning accuracy of the network approach against the traditional technique of using a single reference station was next investigated. This was performed by comparing coordinate estimation of the 13 test points, estimated in real time using the DVRS data, with their estimation using a single reference station using the same set of measurements. For the latter case, the "AlQusies" reference station of the DVRS network was used to ensure consistency of the results. The data used were the archived data of the " AlQusies" station and the internally stored dual-frequency phase data of the rover receiver. The data of the reference and rover receivers were processed in the traditional technique in post mission. The rover points were approximately 9.2 km away from the reference station, therefore processing in a single baseline mode to the cm level of accuracy was feasible in general for the first and third DVRS tests. However, due to the large baseline length encountered, and high PDOP values experienced during the second test, post processing of the collected phase results for the second test was unsuccessful. The average standard deviations of point estimation from data of the first and third tests for the single baseline scenario and the DVRS survey are given in Table 4.5. For the former case, the 2D positioning accuracy was usually less than 2.2 cm, while for the height it was less than 5 cm, and its was 11 cm at most. Comparison of the positioning accuracy in the two estimation scenarios show that the DVRS performed better as all of its standard deviations were smaller than their corresponding values computed by using an single reference station except for the Easting coordinate in the first test. Even in this case, the standard deviations of the two techniques were very close. Thus, it can be concluded that, in general, the DVRS has a better positioning performance than using a single reference station for short baselines. 
To quantify accuracy differences between the two approaches for the tests in hand, discrepancies in coordinate estimation for the 13 points between the DVRS approach and the single reference station processing for test1 and test3 are illustrated in the Figures 4.7 and 4.8, respectively. In general, the differences were less than 5 cm for the planimeteric coordinates and 7.4 cm for the height. The standard deviations estimated from the differences were less than 6 cm for both cases, as given in Table 4.6.  Figure 4.6 Coordinate discrepancies for test1 with single-baseline processing  Figure 4.7 Coordinate discrepancies for test3 with single-baseline processing 
5. Conclusions The performance of the Dubai Virtual Reference System (DVRS) has been investigated as an example of the RTK networks. The system absolute accuracy was first tested by comparing the DVRS estimated coordinates for a set of 13 points with their accurate coordinates, which have been previously determined by a precise surveying using a total station. The 3D (spatial) positioning differences between the two techniques, reflecting the DVRS external accuracy, ranged between 0.81cm and 3.61 cm. The accuracy of relative positioning was tested by studying differences between distances derived from the DVRS estimated point coordinates against their precise values. The differences were within 1 cm on the average for the three tests, with a maximum value of 2.2 cm. The standard deviation of the differences ranged from 1 to 2 cm. For the three DVRS tests, coordinate precision were less than 2 cm for planemeteric coordinates, and less than 3 cm for height determination. Different DVRS surveys to the same test points prove to be consistent in terms of their internal precision of the final output results. This can be statistically interpreted that the different tests are statistically compatible. In addition, the system proves to be reliable and robust particularly in case of failure of one of the reference stations. In this case, positioning accuracy at the cm level was feasible for points that are as far as 25 km away from the nearest reference station. Such a case can not be usually solved at that level of accuracy using the classical single reference RTK approach. This shows that the DVRS REFERENCES
Page 3 of 3
| Previous | |