1 International Symposium on GPS/GNSS October -8, 1. Performance of Research-Based N-RTK Positioning System in ISKANDAR Malaysia Shariff, N. S. M., Musa, T. A., Omar, K., Ses, S. and Abdullah, K. A. UTM-GNSS & Geodynamics (G&G) Research Group, Infocomm Research Alliance (IcRA), Faculty of Geoinformation Science and Engineering, Universiti Teknologi Malaysia (UTM), 8131 UTM Skudai, Johor, Malaysia. Phone: +7-383; Fax: +7-13 suryatishariff@gmail.com Abstract A research-based N-RTK positioning system namely ISKANDARnet which is situated in equatorial region has been established. The network corrections are generated to improve users position accuracy. In order to testify the performance of the network corrections and positioning quality, tests are conducted at various locations within the network coverage. Real-time test results indicate that positioning accuracy is achievable in centimetre-level. Key words Distance-dependent errors, Network corrections, N-RTK. 1. Introduction Over the past few years, GPS networks of Continuously Operating Reference Stations (CORS) have been extensively deployed for range of purposes. Particularly for high accuracy positioning over large area, the CORS network is beneficially being used by the Network-based Real-time Kinematic (N-RTK) positioning technique. This technique utilizes carrier phase-based measurements from CORS network and models the distance-dependent errors such as ionospheric, tropospheric and orbital errors that respect to rover. A direct result of modeling the distancedependent errors is the ability to improve the resolution of ambiguities, which are important to form high-precision RTK survey [1]. However, in area of severe atmospheric activities especially in equatorial region, there is a challenge to reliably model the distance-dependent error []. A research-based N-RTK system namely ISKANDARnet, which is located in equatorial area has been developed by Universiti Teknologi Malaysia (UTM) and collaboration with University of New South Wales (UNSW). The establishment of ISKANDARnet has provided a good platform to various studies such as atmosphere, meteorology and positioning in equatorial area. This paper generally describes the ISKANDARnet system in term of architecture and operational as well as evaluate its performance on N-RTK corrections and positioning results. The assessment of network corrections are tested against without apply network corrections. Tests are also carried out in various locations of ISKANDARnet area to perform N-RTK positioning results. Significant experience in N-RTK have yield to discuss on extension of ISKANDARnet service, include network integrity and reverse-rtk.. Architecture of ISKANDARnet Presently, ISKANDARnet consist of three CORS (see Fig.1); ISKANDARnet1 (ISK1) is situated at Faculty of Geoinformation Science and Engineering, UTM; ISKANDARnet (ISK) at the Port of Tanjung Pelepas (PTP), Gelang Patah; and ISKANDARnet3 (ISK3) at Kolej Komuniti, Pasir Gudang that covers metro-area of Iskandar Malaysia (http://www.iskandarmalaysia.com.my/). Fig. 1: Distribution of ISKANDARnet Reference Stations. (http://www.fksg.utm.my/iskandarnet/about.html). The ISKANDARnet system comprises of three main components which are the CORS, a control center and user applications (see Fig.). Beside to aim for N-RTK service, ISKANDARnet also provides multi-functional services such as single-based RTK, Differential GPS (DGPS) and Receiver Independent Exchange Format (RINEX) data download. 1
1 International Symposium on GPS/GNSS October -8, 1. Table 1: Distance Test Points from Reference Stations. Points ISK1 ISK ISK3 PT1 1.9 km km 1.8 km PT.8 km 1. km 7.3 km PT3 1.8 km 9. km 1.8 km PT.1 km 3.8 km 7. km PT 17.7 km.9 km 17.9 km PT 19. km.3 km 37. km PT7 1. km 11. km 3. km In order to evaluate performance of ISKANDARnet N-RTK system, three tests were carried out as in the following sub-sections: Fig. : Architecture of ISKANDARnet. For ISKANDARnet implementation, the N-RTK processing software generally has four crucial steps which are master-to-reference station processing, generating network corrections, corrections distribution, and user processing. In master-to-reference station processing stage, one of reference station (usually the nearest) will be selected as a master station to resolve the network ambiguities and generate the network residuals. The use of fixed network residuals will ensure that high-quality network corrections can be generated through the interpolation process []. Biases for any user in network area are interpolated based on approximate user location. Next, network corrections can be distributed via Virtual Reference Station (VRS) method [3], [] and [] to facilitate user-side processing. 3. ISKANDARnet N-RTK Performance A few observation campaigns have been carried out from th - 7 th May, th - th Jun, 1 th and 1 th August 1. There are seven proposed user locations (see Fig.3). Table 1 shows the distance of each test point from reference station ISK1, ISK and ISK3. 3.1. N-RTK vs Single-based RTK This test evaluates the performance of N-RTK (i.e. with network corrections) relative to the single-based RTK (i.e. without network corrections) approach. The RTKLIB-rtk post processing software [] has been used to process all test points with the nearest reference station (see table 1). Static observation was also carried out during the campaign to obtain known coordinates of all points. Table : Position Errors - with (w) and without (w/o) Network Corrections. (a) PT1 w/o.7.7.8 -.8..3 -.91 1.9 3.8 w 1..8 1.19 -.9.1.18 1. 1..17 3.9. 33.8 (b) PT w/o 1.3.77 1.81-1.18.7 1.37 -. 1.7 3. w.7 1. 1. -1. 1.1 1. -1.1.17.9. -.8-1.1 (c) PT3 w/o -. 1.33 1..9.77 1. -11..31 13.19 w -.8 1. 1.9 1.77. 1.88 -.3 3.7. 11. -1.. (d) PT w/o.3 1.. 1.8 1.8 1.99 -.31.7.1 w 1.99 1..9-1. 1.8 1.38 -.7.7.9 1. 3.7 8. (e) PT w/o 7.3.7 7.3 8..8 8.3 1.1.7 1.1 w -.1.39. -.83.9 1.1.9.8 3. 7.1 87. 7. Fig. 3: Observation Campaign Area.
1 International Symposium on GPS/GNSS October -8, 1. (f) PT Position Error at PT.8.8 79..9 1..3 1. -.8.9 -..3.91. -. 8.31.3 8.33 w/o -1.1.93 1.8 w 77.8 (g) PT7 -.1.91 -.3.83.1.98.7 -.88 1.7 -. 3.13.9 3.17 7.9 8 1 1 1 1 3 1 1 3 1 1 3-1 [cm] - 1 3 1 3 1 3 - - (c) Position Error at PT 18-1 1 1 1 1 1-1 -1 (d) - - 8 1 1 1 1 18 8 1 1 1 1 18 1 [cm] 1 Position Error at PT3 [cm] 1 (b) Position Error at PT1 1 - - Tables (a) - (g) show numerical results of position errors at each test point over its known coordinates. It can be noticed that the position errors of N-RTK approach constantly provide centimetre level - in terms of average, standard deviation and values. These tables also indicate percentage in of N-RTK is much improved over the single-based approach, except for component of PT and PT3, and component of PT and PT. The test point of PT shows significant improvement up to 7.1%, 87.% and 7% in, and components, respectively. Since the distance of PT is over than 17 km from ISK1 (see Table 1), this result obviously explains that the network correction effectively reduces the effect of distancedependent errors compared to the single-based approach. Figures (a) - (g) provide the time series of position errors for both the N-RTK (green line) and single-based RTK (blue line). All figures show discrepancies in position errors of N-RTK in the, and components are fairly consistent at the zero value. However, there are a few drop-off in the position errors of the N-RTK approach at certain points as shown in Figures (b), (d) and (f). This could be the results of poor quality of the network corrections. Further inspection has found that these points (PT, PT and PT see Fig. 3) were located at the corner of the ISKANDARnet triangle. This result support the finding by [7] indicates that the performance of the network corrections decrease as the N-RTK user is moved far away from the center of the network. - [cm] -7. 1. 7.3 w/o 1..7 1.7 w 77. -1 (a) 3
[cm] d Up [cm] d Up [cm] 1 International Symposium on GPS/GNSS October -8, 1. 1 Position Error at PT -1 8 1 1 1-1 8 1 1 1 8 1 1 1-1 (e) Position Error at PT 1 1 3.. Real-Time Test of ISKANDARnet The ISKANDARnet was also evaluated in real-time mode by collecting 1 epochs of N-RTK positioning at each test point (see Fig.3). The N-RTK coordinates are compared with the known coordinate (as obtained from static observation in section 3.1). Fig. Fig.11 show scatter plot of position errors in (dn), (de) and. The blue dots indicate the position errors that relative to the known position (denoted as red dot). From these figures, it can be seen that small variations of position errors in horizontal component with standard deviation achievable is less than 1.3 cm and up to.1 cm in vertical component. These figures also show that the position error distributions are not exactly close to the known point. Perhaps, it is due to known position is offset from the true value. However, the accuracy achieved is reasonable, according to average values are in the range of centimeter level. 1 8 Position Error of PT1-1 1-1 1 (f) - - - d North - - d East Fig. : Scatter Plot of Position Error for PT1. 1 Position Error at PT7 Position Error of PT -1 1 1 3 3 1-1 1 3 3 1-1 1 1 3 3 (g) Fig. : Position Errors in, and Component of the N-RTK (green) and single-based RTK (blue) at All Test Points. - -1 - - d North - - d East Fig. : Scatter Plot of Position Error for PT.
d Up d Up d Up d Up d Up 1 International Symposium on GPS/GNSS October -8, 1. Position Error of PT3 Position Error of PT7 1 1 8 - -1 d North - - - d East 8 d North - - -1 - d East Fig. 7: Scatter Plot of Position Error for PT3. Fig. 11: Scatter Plot of Position Error for PT7. 1 - -1 1 Position Error of PT - d North - - d East Fig. 8: Scatter Plot of Position Error for PT. Position Error of PT 3.3 ISKANDARnet vs MyRTKnet Test of ISKANDARnet N-RTK performance was also carried out by comparing with the commercially available national N-RTK system, Malaysia Real-Time Kinematic GNSS Network, known as MyRTKnet [8]. A total of 1 epochs of fixed positioning solutions were recorded for both N-RTK systems, thus average it to be compared. It is assumed that there is no rapid atmospheric change since there is only short delay of about 1 minutes between ISKANDARnet and MyRTKnet observations. Table 3 shows the coordinates average and standard deviation from ISKANDARnet as compared to MyRTKnet at PT1, PT and PT7. Overall, the average coordinates from both N-RTK systems differ only in centimeter level with maximum of. cm for the component. The results also shows small coordinates deviation between the two network systems as indicated in standard deviation values in Table 3. Hence, the ISKANDARnet positioning results is reasonably compatible with the MyRTKnet. - -1 3 3 1 1-1 d North -1 - d East Fig. 9: Scatter Plot of Position Error for PT. Table 3: Position rage and Standard Deviation for NRTK of ISKANDARnet Compared to MyRTKnet. rage Standard Deviation PT1-1. -.....7 PT.3-1.8...3.3 PT7-1.3 -. -. 1. 1.8 3.8 1 - -1 d North Position Error of PT - d East Fig. 1: Scatter Plot of Position Error for PT.. Overview of Future Works Uncertainty of N-RTK environment possibly provides unreliable network corrections. It is becomes critical when user is not noticeable of any malfunction that could affect their position accuracy. For instance, section 3.1 has shown that N-RTK has accuracy drop-off even though in fixed positioning solution. Thus, to ameliorate potential data quality concerns, the integrity monitoring is crucial to be undertaken by giving a timely warning to users when large position errors occur [9]. In future, ISKANDARnet is inspired to adopt integrity monitoring system which detects the status of reference stations, atmosphere conditions within network coverage, and availability of real-time data streams thus to inform users about the quality and fitnessfor-purpose of N-RTK positioning results.
1 International Symposium on GPS/GNSS October -8, 1. Such of quality indicator is not only potential for N- RTK system, but become increasingly attractive for advanced positioning method called reverse-rtk. The reverse-rtk is a server-based approach, which service provider can exercises control over the generated products and place a true commercial value [1] by providing a value-added product to the user. However, it can be expected that the preparation towards implementing the reverse RTK will involves with great data processing and management as well as communication links.. Concluding Remarks The performance of ISKANDARnet has been demonstrated in this study. The observation campaigns indicate that the network correction yields significant improvements in position errors relative to single-based approach due to the capability of N-RTK to model the effects of distance-dependent errors. Furthermore, reasonable accuracy of centimeter-level is achieved in realtime test of the system. The positioning result of ISKANDARnet is also found compatible within centimeter level with the commercial MyRTKnet system. Currently, implementation of network integrity monitoring for ISKANDARnet is in progress. Further work will also include a development of reverse-rtk system for ISKANDARnet. [] Takasu, T. (1). RTKLIB ver.... Manual. [7] Dao, D., Alves, P. and Lachapelle, G. (). Performance Evaluation of Multiple Reference Station GPS RTK for a Medium Scale Network. Journal of Global Positioning Systems. 3 (1), 173-18. [8] Jamil, H., Mohamed, A. and Chang, D. (1). The Malaysia Real-Time Kinematic GNSS Network (MyRTKnet) in 1 and Beyond. FIG Congress 1, 11-1 April, Sydney, Australia. [9] Chen, W., Hu, C., Ding, X., Chen, Y. and Kwok, S. (). Critical Issues on GPS RTK Operation using Hong Kong GPS Active Network. Journal of Geospatial Engineering. (1), 31-. [1] Rizos, C. (7). Alternatives to current GPS-RTK Services and Some Implications for CORS Infrastructure and Operations. GPS Solution 11: 11-17. References [1] Hu, G. R., Khoo, H. S., Goh, P. C. and Law, C. L. (3). Development and Assesment of GPS Virtual Refernce Stations for RTK Positioning. Journal of Geodesy. 77, 9-3. [] Musa, T. A., Lim, S., Yan, T. and Rizos, C. (). Mitigation of Distance-Dependent Error for GPS Network Positioning. International Global Navigation Satellite Systems Society IGNSS Symposium, 17-1 July, Queensland, Australia. [3] Landau, H., Vollath, U. and Chen, X.,. Virtual Reference Station Systems, Journal of Global Positioning Syste. Vol. 1(),137-13. [] Wanninger L () Virtual Reference Stations for Centimeter-Level Kinematic Positioning. ION GPS, September 7,, Portland, Oregon, pp 1 17. [] Rizos, C. and Han, S. (3). Reference Station Network-based RTK Systems-Concepts and Progress. Wuhan University Journal of Natural Sciences. 8 (), -7.