Does GNSS outperform GPS in Geodetic Applications?

Transcription

1 Contribution to Conference Proceedings of POSITIONs 008, Dresden Does GNSS outperform GPS in Geodetic Applications? Asim Bilajbegović, Prof. Dr.-Ing. 1, Rüdiger Lehmann, Prof. Dr.-Ing. 1, University of Applied Sciences Dresden 1 bilajbegovic@htw-dresden.de r.lehmann@htw-dresden.de Summary In decision-making process, what kind of GPS equipment to purchase, one faces dilemma, to take eir GNSS (=GPS+GLONASS) or GPS receivers only. In case of full completeness of GLONASS satellite system, this dilemma would certainly not exist. The solution to this dilemma is given for a constellation of 14 operational GLONASS satellites. Due to short operational period of se satellites (for example GLONASS-M only 5 years), and not launching new ones, in this moment (July 008), re are only 16 satellites operational. In our research work we used about 5 RTK measurements obtained with both GPS only and GNSS receivers. We will show how answer to dilemma depends on obstruction of horizon at station. Besides that, initialisation time of both systems has been investigated on basis of about 480 measurements, using rover s antenna with metal cover, during intervals of 0, 5, and 5 seconds. Finally, accuracy has been investigated and compared to accuracy and redundancy of GPS and GNSS RTK measurements given by manufacturer. 1 Introduction The Tables 1 and show simplest interpretation and overview of nowadays and futures developing satellite positioning systems. Table 1: The overview of GPS satellites and irs signals, by Reaser (006) Signal Year of Satellite L1 L1 L1 M L1C L LC L M L5 launching block C/A P(Y) P(Y) or planned launching IIR IIR-M Sept th satellite IIF March 008 IIIA Jun 013

2 Table : Overview of GLONASS satellites and signals used, by Dvorkin et al (006) Average Year of Satellite Signal / frequency of launching Total number of satellite s first satellites life span satellite L1 L L GLONASS years GLONASS-M 5 years Middle of GLONASS-K 7 years Planned for middle of March Planned Full Planned constellatio 009 n 4 Both Tables clearly show planned modernisations and transitions from two to three carriers for both systems. Three frequencies will improve accuracy, reliability and initialization time of rover. For getting information of accuracy, precision and economy of modern GPS and GNSS (GPS+GLONASS) receivers, it was necessary to take a right method of measuring, testing network, mono and hybrid satellite receivers. Since 1994, RTK method of measuring has been developing. After SAPOS network of permanent stations has been established in Germany, in about 95% of cases RTK method is used in practice. In geodetic applications, highly precise real time positioning service (HEPS) is favoured. Therefore, for this investigation, RTK method has been used throughout. The testing network of University of Applied Sciences Dresden has been used. This network is situated in flood planes of river Elbe in centre of Dresden, Fig.1. The horizontal uncertainty of points in test network is σ L = 10 mm, and for heights σ h =5mm.

3 Fig. 1: Test network of University of Applied Sciences Dresden Trimble R8 GPS and R8 GNSS receivers were used. During field test measurements, need for new construction solution occurred, which would enable transport of two receivers by one person, Fig., 3 and 6. Primarily due to influence of ionosphere, troposphere and identical constellation of satellites, measuring had to be carried out consecutively in time. Therefore, one carrier pole was used for both antennas, for both receivers and for two portable field computers ACU(Advanced Control Unit), Fig., 3 and 6. Fig. : Transport over short distances Fig. 3: Transport over longer distances In German federal state of Saxony, permanent stations are not equipped with GNSS receivers. Therefore, our own permanent station was used. During this investigation corrections were transmitted by our own radio transmitter and using GPRS via NTRIP protocol and RTCM 3.0 format.

4 Short description of equipment The modern Trimble R8 GPS and R8 GNSS receivers were used in this investigation. Trimble R8 GPS is a dual-frequency, 4 channels GPS receiver, with integrated GPS antenna and 450 MHz radio-transmitter, Fig. 4. This system enables recording satellite signals at low elevations. It enables completely wireless Bluetooth -communications between receiver and control unit (ACU). Besides phase observations on L1 und L carriers, code on C/A on L1 carrier, P-code on L1 and L carriers, using this technology LC signal can also be recorded. (The first satellite with LC signal was launch on September 5, 005). Fig. 4: Trimble R8 GPS-receiver Fig. 5: Trimble R8 GNSS-receiver Maximal distance of transmitting RTK-correction from base station, in accordance with firmware declaration, is 3-5 km by a transmitter power of 0.5 W. This investigation showed problems at much shorter distances of 1.6 km. All Trimble receivers can be used as rover as well as base station. The initialisation time of rover is 10 seconds seconds/km for distances up to 30 km. The Trimble R8 GNSS-receiver is a multi-channel and multi-frequencies GNSS-system. The number of channel is 7, which enable recording L1 C/A-code, P-code on L1 and L carriers, LC code, phase measurements on L1, L and L5 carriers, and recording GLONASS signals as well, Fig. 5. A new RTK-engine (Trimble Maxwell TM Custom Survey GNSSchip) enables very fast initialization of rover. According to firmware declarations, an initialisation takes less than 10 seconds. In chapter 5, se times were investigated more closely.

5 3 Test networks and analysis of measurements The test network of University of Applied Sciences Dresden was used for this investigation. Horizontal position and height accuracy of network is better than 10 and 5 mm, respectively. The network consists of 38 points with no or low obstructions of horizon, 48 points with medium obstructions up to 35% and 40 points show high obstructions of horizon above 35%. The measurements have been done in two sessions using each rover, Table 3, 4, 5 and 6. For elimination of outliers, it was necessary to define tolerances for differences between observed coordinates and known coordinates of test network. The rover s antenna height was only m, for reason of fast changes of GPS and GNSS receivers during measuring at each point, Fig. 6. Sensitivity of circular level on carrying pole is 8' for mm of bulb shift, what introduces a centring error of 4. mm in horizontal plane and only a negligible error in antenna height. On base of mentioned errors in uncertainty of coordinates of test network, horizontal (10mm+1mm/km) and vertical (0mm+1mm/km) uncertainty, specified by manufacturer, maximal distance between rover and base station of 3.6 km, and propagation of errors, expected standard deviations along side x and y axis and heights results as follows: σ = mm and σ = = 4. mm, (1) x, y = 6 h 1 For a probability of 99% and χ distribution adequate tolerance deviations are:, = = 3.4 mm; = = 6.1 mm = mm. x y h L = These tolerance s were used for filtering outliers. Table 3: List of results for points, without or with low obstructions of horizon; in two sessions, 76 attempts of measuring using each receiver Sum of Receiver measurements with measurements with unsuccessful R8 standard deviation standard deviations measurements advantage within tolerance out of tolerance of R8 GNSS %.63%.63% GNSS GPS % 1 1.3% % Advantage of GNSS R8 in percents with sign % -1.31% +7.90% %

6 Table 4: List of results for points, with medium obstructions of horizon; in two sessions, 96 attempts of measuring using each receiver Sum of Receiver measurements with measurements with unsuccessful R8 standard deviation standard deviations measurements advantage within tolerance out of tolerance of R8 GNSS % % % GNSS GPS % % % Advantage of GNSS R8 in percents with sign % 0.00% +7.9% % Table 5: List of results for points, with high obstructions of horizon; in two sessions, 80 attempts for measuring using each receiver Sum of Receiver measurements with measurements with unsuccessful R8 standard deviation standard deviations measurements advantage within tolerance out of tolerance of R8 GNSS % % % GNSS GPS % % 7.50% Advantage of GNSS R8 in percents with sign % % -3.75% +1.51%

7 Table 6: Recapitulation of all measurements carried out (points with low, medium and high obstructions of horizon); in two sessions 5 attempts measuring using each receiver Sum of Receiver measurements with measurements with unsuccessful R8 standard deviation standard deviations measurements advantage within tolerance out of tolerance of R8 GNSS % % % GNSS GPS % % % Advantage of GNSS R8 in percents with sign % +.78% +3.97% % Table 3, 4, 5 and 6 are reporting by mselves, and no comments required. The advantage is obvious in redundancy of hybrid R8 GNSS-receivers, for all kind of obstructions of horizon. It makes up 13.5%, shown in last column of Table 6. After eliminating measurements with outliers, standard deviations of single measurements calculated, depending on rate of obstruction of horizon, alongside of coordinate axis x, y, h, as well as, in horizontal plane and in heights, Table 7. Table 7: Standard deviations of a single measurement, depending on rating of obstruction of horizon. Standard deviations for R8 GNSS Standard deviations for R8 GPS Obstructions horizontal horizontal y x heights y x heights of plane plane [mm] [mm] [mm] [mm] [mm] [mm] horizon [mm] [mm] low medium high From Table 7, one can see advantage in horizontal position accuracy of coordinates stated by R8 GNSS-rover for all cases of horizon obstructions. The same statements can be made for accuracy of determining heights. It is important to note that all measurements were testing for normal distributions.

8 Fig. 6: Pole with both rovers, both ACUs and transporting bike. 4 Investigations of accuracy declared by manufacturer The task was to investigate accuracy declared by manufacturer, for horizontalpositional and height precision (accuracy) as function of distance between rover and base station, (10mm+1ppm and 0mm+1ppm). As longest distance between rover and base station was 3.6 km it follows, that achieved positional-horizontal precision should be 13.6 mm and standard deviation for heights should be 3.6mm. Standard deviations from Table 7 calculated from residuals of known coordinate of known points and point s coordinates determinded by RTK measurements. It means, se standard deviations are biased by errors of coordinates of known points. They totalize, for horizontal plane 10 mm, alongside axis x and y up to 7 mm, and for heights (levelling) 5 mm. Using Table 7 achieved precision of measurements can be evaluated by following formulas: where p n p n s L = ( sl) 10 i sh = ( sh ) 5, () n s L - Standard deviation in horizontal plane, Table 7 p sl - Positional horizontal precision, Table 8 n sh - Standard deviation of height, Table 7, and p sh - Precision of measuring height, Table 9.

9 Table 8: Test of horizontal positional precision in accordance to manufacturer declaration obstruction of horizon Positional horizontal precision in accordance manufacturer declaration 13,6 p sl u [mm] R8 GNSS Testing Significant limits of testing mm p sl u [mm] R8 GPS Testing Significant limits of testing low medium high Table 9: Test of precision of height in accordance to manufacturer declaration obstruction of horizon Precision of height in accordance to manufacturer declaration 3,6 mm p sh u [mm] R8 GNSS Testing χ Significant limits of testing p sh u [mm] R8 GPS Testing χ Significant limits of testing low medium high The test of standard deviation was implemented as follows: The Null hyposis is H : s = σ 0 0, Testing Where: χ k s = σ k=n-1: redundant number of measurements, s: empirical standard deviation, and σ : standard deviation by manufacturer specification. 0, The significant limit of testing χ can be taken from table for χ distribution, for probability of 99% and variable number of degrees of freedom k, Tables 8 and 9.

10 If χ < χ n null hyposes is acceptable. From Table 8 one can see, that both receivers fulfilling declared horizontal positional precision. On contrary, declarer precision of determining heights was not reached by any receiver, Tab. 9. This can be partially explained by different influence of multipath-effects for different antenna type used during determination of coordinates of test network, Bilajbegović et al (007), Wanninger et al (006). Position precision, obtained by GNSS receiver is, in average, for mm better (or 14%), but, precision of determination of heights is worse by mm (or 0.9-.%). 5 Investigation of initialisation time of receivers With a view to investigation of receiver s initialisation time a reference station was placed at a distance of about 100 m from rover. The rover s antennas were separated by about m and were covered by metal cover, for period of 0.5, and 5 seconds, Fig. 7. The measuring were implemented successively, one after anor one, for both receivers. 80 measurements were taken, using both systems, and for each period of covering of antenna, or in total, 480 measurements. The initialisation measurements were carried out at two days. For sake of elimination of eventual outliers, coordinates of rovers were determined from 1000 measurements during first day, but, from 4000 measurements during second day. The recorded measurements enabled to readout lost initialisation timeintervals and re-initialisation time-intervals of receivers. Analysing resulting measurements, it was possible to identify two intervals of initialisation, as follows: after lost of initialisation due to high standard deviations and interval time of initialisation after lost of initialisation in consequence of covering antenna. Selected investigating results are displayed in Table 10. Fig.7: Investigation of initialisation times of receivers

11 Table10: Overview of intervals of initialisations of rover receiver Time-interval after initialisation lost because of high standard deviations Time-interval after initialisation lost because (antenna covering) Generally, average time-interval of initialisation Time interval after covering antenna for about 0,5 seconds Time interval after covering antenna for about seconds Time interval after covering antenna for about 5 seconds R8 GNSS [seconds] Declared time-interval of initialisations R8 GPS [seconds] Declared time-interval of initialisation 13.9 < 10 s s 8.0 < 10 s s 8. < 10 s s 5. < 10 s s 6.9 < 10 s s 1.5 < 10 s s Manufacturer s declaration for time-intervals of initialisation for R8 GNSS is <10 seconds, but for R8 GPS receivers 10 s +0,5 s/km (up to 30 km distances from reference station). Obviously, in accordance to Table 10, time-interval of initialisation is a function of lasting of covering of antenna, and it is shorter for shorter time of covering antenna. Beside of that, average time of initialisation of R8 GNSSreceiver is shorter, an in average, is tree times shorter relative to R8 GPS-receiver, and is even shorter than declared manufacturer time. Time-interval of initialisation for R8 GPSreceiver is longer for about,5 times than manufacturer declared time. 6 Conclusions The investigation described in this paper shows that for constellation of 14 GLONASS satellites: Hybrid R8 GNSS-receiver are more reliable relating to R8 GPS, for measurements on points with horizon obstructions: low, medium and high ones, expressed in percents, it is better by 13%,

12 The horizontal positional standard deviation is better by about 14% than for R8 GPSreceiver, The accuracy for heights is same as for R8 GPS, The initialisation time-interval is.5 times shorter than for R8 GPS, The initialisation time-interval is a little shorter than in manufacturer s specification, The both systems R8 (GNSS and GPS) yield horizontal precision specified but not for heights, Receiver R8 GPS has almost.5 times longer initialisation time relating to manufacturer s specification, Using hybrid system R8 GNSS one can determine about 70% points in cities area, whereas, by receiver R8 GPS about 63% points. It s worth noting that all statements above are valid for a constellation of 9 GPS and 14 GLONASS satellites and for area of test network of University of Applied Sciences in Dresden. It varies depending on number of available satellites and measurement location. 7 References Bilajbegović, A., Vierus, M. (007): Untersuchung der Multipath-Effekte verschiedener GPS- Antennentypen und ihrer Einflüsse auf die Genauigkeit der Koordinatenbestimmung. Allgemeine Vermessungsnachrichten 1/007, pp Dvorkin, V., Karutin, S. (006): GLONASS: Current status and perspectives. Allsat open conference, Hannover, Reaser, C. R. (006): Navstar Global Positioning System. Allsat open Conference, Hannover, Wanninger, L., Rost, Ch., Hartlieb, G., Köhr, M. (006): Zur Problematik des Antennenwechsels auf GNSS-Referenzstationen. Zeitschrift für Vermessungswesen 4/006, pp