EVALUATION OF ABSOLUTE AND RELATIVE CARRIER PHASE POSITIONING USING OBSERVATIONS FROM NAVIGATION-GRADE U-BLOX 6T RECEIVER

Transcription

1 31 August - 2 September 211, Copenhagen, Denmark. EVALUATION OF ABSOLUTE AND RELATIVE CARRIER PHASE POSITIONING USING OBSERVATIONS FROM NAVIGATION-GRADE U-BLOX 6T RECEIVER Constantin-Octavian ANDREI 1, Martin VERMEER 1, Heidi KUUSNIEMI 2, and Hannu KOIVULA 2 1 Department of Surveying, Aalto University, Otakaari 1 Y, 76 AALTO, Finland, octavian.andrei@aalto.fi 2 Finnish Geodetic Institute, Geodeetinrinne 2, 2431 Masala, Finland, firstname.surname@fgi.fi ABSTRACT Geodetic and surveying community has traditionally used geodetic-grade receivers and precise calibrated antennas for precise positioning. Although they allow users to obtain high accuracy levels, geodetic-grade equipment is still expensive. However, GNSS receiver and antenna technology has been under intense development over recent years. Therefore, high-sensitivity navigation-grade receivers may be used to achieve cost-effective precise positioning. The requirement for these types of receivers is to output carrier phase measurements, the most accurate measurements provided by a GNSS system. This paper provides a preliminary evaluation of the positioning performance of high-sensitivity navigation grade u-blox LEA-6T receiver. Several tests were conducted to analyze the performance and the achievable accuracy. The results demonstrate cm-level positioning accuracy can be obtained (in certain conditions) with highsensitivity navigation-grade u-blox LEA-6T receivers when they can output carrier phase data. This level of accuracy opens market for various applications, such as land, marine or aerial surveying, structural monitoring or early warning systems. Key words: GNSS, LEA-6T, RTKLIB, low-cost precise positioning, carrier-phase measurement; navigationgrade receiver. 1. INTRODUCTION Global Navigation Satellite System (GNSS) allows positioning with accuracy ranging from a few tens of meters down to a few millimeters. The accuracy level depends on the GNSS equipment, the positioning technique and the surrounding environment. Traditionally, geodeticgrade receivers and precise calibrated antennas have been used for accurate positioning. Although they allow users to obtain high accuracy levels, geodetic-grade receivers are still expensive, ranging from 1 to 3e[18]. On the other hand, GNSS receiver and antenna technology has been under intense development over recent years. As a result, precise positioning has also become possible using high-sensitivity (HS) navigationgrade GNSS receivers. The navigation-grade receivers are attractive because they are significantly cheaper (1 to e) than the geodetic-grade receivers. In addition, navigation-grade receivers are up to 3 db more sensitive [3, 16], which enables higher GNSS signal availability in places, such as urban canyons or even indoor. The main requirement for these navigation-grade receivers is to output carrier phase measurements, the most accurate measurements provided by a GNSS system. Over the recent years, several studies have been conducted to investigate the potential of improving the positioning accuracy of low-cost receivers [, 8, 4, 11, 12, 1]. Although the reported results were positive, the manufacturers did not find them attractive enough to further develop navigation-grade receivers that output carrier phase data. As a result, there are few such type of receivers currently on the market. Moreover, some of the manufacturers (e.g., Garmin) do not provide official interface for raw phase data, while others (e.g., SiRF) document only internally their phase data, with no access for potential users [6]. On the other hand, very few other manufacturers, such as u-blox, give access to their receiver technical documentation. Having this said, this paper presents the first results obtained using the latest u-blox receiver, namely LEA-6T receiver, which supports raw phase data output. The objective of this study is to analyze the performance of both the receiver and the system (receiver + antenna) as well as the potential achievable accuracy through several field tests. 2. METHODS AND EQUIPMENTS 2.1. GNSS positioning techniques Global Navigation Satellite System (GNSS) technology has been adopted by various number of industries and business because it can provide accurate position, velocity and time information to anyone who owns a GNSS receiver. The position information can be obtained through several positioning methods mainly depending on the accuracy requirement and the type of the GNSS receiver. 1

2 31 August - 2 September 211, Copenhagen, Denmark. For many users and applications, GNSS provides absolute positioning information, based on code-only or phase-smoothed code observations. The absolute positioning has a limited accuracy due to the various GNSS error sources. However, for many other applications, this level of accuracy is insufficient. Combining observations from a minimum of two receivers results in highly improved relative coordinates between the receivers. A common way to improve accuracy, availability, and integrity is using differential positioning. In differential positioning, a roving receiver receives corrections to the measured pseudoranges from a base station and performs point positioning with corrected values. The use of corrected pseudoranges improves the position accuracy with respect to the base station. To further improve the accuracy, relative positioning can be used. In relative positioning, the user computes the coordinates of an unknown point with respect to a known point by determining the vector between the two points. The base station transmits the time-tagged raw data to the roving receiver. At the rover, these data are differenced with the corresponding rover data. In the end, the rover receiver estimates its own position relative to the reference receiver. Relative positioning can be performed with code or phase data. However, only the phase-based approach is considered because the solutions based on phase observations are far more accurate than the ones obtained using code observations. Therefore, high positioning accuracy is achieved. There are different ways of relative positioning, such as static positioning, real-time kinematic (RTK) or network RTK (e.g. with the virtual reference station method, VRS). This study focuses mainly on all these relative carrier phase positioning techniques. Additionally, this study involved various hardware and software equipment for collecting the field data. A short overview is given below Hardware The u-blox 6 precision timing evaluation kit EVK-6T package represents the main focus of this investigation. The package includes a compact box (74 x 4 x 24 mm) that contains the LEA-6T module (Fig. 1a), a u-blox ANN-MS low-cost patch active antenna (Fig. 1b), the USB interface cable and a CD-ROM containing: evaluation software, USB-driver software and technical documentation. The package becomes attractive because does not require an external power supply; EVK-6T comes with a built-in USB interface for both power supply and high-speed data transfer. In addition, the EVK-6T includes the u-center, an interactive tool for configuration, testing, visualization and data analysis of u-blox receivers. As a result, the EVK-6T can be used with a notebook or PDA, making it suitable for use in both indoor and outdoor locations. The main feature of the LEA-6T comes from the fact that it supports raw data output at an update rate of Hz. According to the technical documentation, the UBX-RXM- RAW message includes carrier phase with half-cycle am- Figure 1. Various hardware equipment used during field operation: navigation-grade receiver (a), low-cost patch antenna (b), signal splitter (c), and connectors (d). biguity resolved, code phase and Doppler measurements, which can be used in external applications that offer precision positioning, real-time kinematics (RTK) and attitude sensing. In addition to u-blox equipment, geodetic type hardware was used for data collection. Two high-end dualfrequency GNSS receivers (i.e, Ashtech Z-XII and Leica GPS 12) and two high-quality antennas (ASH7228D and LEIAX122) were used to assess, compare, and validate the performance of the LEA-6T system. These receivers provide code and phase observations both on L1 and L2 frequencies that allow to compute high accurate positioning solutions. In other tests, two receivers were connected to the same antenna with the help of a GNSS signal splitter (Fig. 1c). The two-way signal splitter has one port DC pass (providing a means to power an active GNSS antenna) and other port DC blocked and is furnished with a 2 ohm internal load to simulate antenna preamp current draw, thus preventing the GNSS receiver from sensing antenna fault. Additionally, TNC to SMA and TNC to N-type connectors are used (Fig. 1d) Software Many of the open-source, commercial or scientific GNSS processing software programs allow to work with either proprietary formats or the standard Receiver INdependent EXchange (RINEX) format. As given in section 2.2, LEA-6T outputs both code and phase data in a binary format (i.e., ubx proprietary format). The resulted ubx file can be converted into RINEX format using, for example, TEQC freeware software developed and maintained by UNAVCO [2, 17]. TEQC utility (i.e., version 21Oct21) can be downloaded freely from the Internet and is an excellent tool for format translation, data editing and quality-check before post-processing. After RINEX conversion, the data collection is imported into a processing software. Over the last decade, many organizations have developed various processing packages. These packages are mainly used to process GNSS data from small or large networks and very often pricy. In this research, we use RTKLIB open-source program li- 2

3 31 August - 2 September 211, Copenhagen, Denmark. East [mm] North [mm] Figure 2. Short-baseline setup brary for GNSS standard and precise positioning [13, 1]. RTKLIB is a compact and portable program library to provide several application programs for RTK-GNSS applications, such as RTKPOST (post-processing analysis) or RTKNAVI (real-time positioning). The library implements fundamental navigation functions and carrierbased relative positioning algorithms [14]. The integer double-difference ambiguities are determined by the LAMBDA (Least-squares AMBiguity Decorrelation Adjustment) method [1]. This study exploits RTKLIB version [9]. 3. TESTS AND RESULTS Various field tests were conducted in the Otaniemi campus of Aalto University, Helsinki, in order to investigate the performance and accuracy of the u-blox 6T system. Two LEA-6T receivers were deployed on the roof of the Department of Surveying building to measure zeroand short ( 3m) baseline. The data collection covered the same time interval, from 9:3: to 12:: (GPSTime), on June 2th, 211 (zero-baseline) and June 22nd, 211 (short-baseline). The observation data were collected at a 1 Hz frequency, with an elevation cut-off angle of degrees. All the observations were stored on a PC. All tests were static, but the data were processed in a kinematic mode, which means that each epoch has one baseline result calculated for the corresponding observation epoch Zero-baseline test In a zero-baseline test, two receivers collect data from the same antenna by using a signal splitter. In this test, all common error sources, such as satellite orbits and clocks, atmosphere and multipath, are eliminated in the baseline processing. As a result, the zero-baseline test allows receiver performance investigation and gives an idea of the observation noise characteristics. Fig. 3 illustrates the zero-baseline positioning results. The baseline compo- Up [mm] Epochs Figure 3. Time series of kinematic positioning for u-blox LEA-6T zero-baseline. North [mm] East [mm] Up [mm] Epochs Figure 4. Time series of kinematic positioning for u-blox LEA-6T short-baseline (the coordinates have been offset by their individual means). nents are expressed in North, East and Up local topocentric coordinate system Short-baseline test In the short-baseline test, the receivers from the same manufacturers in combination with individual antennas are deployed at approximately same height and at a distance of approximately 3 meters from each other as shown in Fig. 2. Because of the short distance between receivers, the satellite- and atmosphere-dependent errors are canceled in the baseline processing. However, the multipath effects are not eliminated. This test evaluates the overall performance of the full system, receiver and antenna. Fig. 4 depicts the short-baseline positioning results in North, East and Up coordinates. The coordinates have been offset by their individual means. The results demonstrate centimeter level positioning precision can be achieved when simultaneously using LEA-6T receiver as reference and rover. Hence, this level of precision is achieved after a convergence period, which for this test was 71 seconds. 3

4 In Proceedings of 3rd International Colloquium on Scientific and Fundamental Aspects of the Galileo Programme, 31 August - 2 September 211, Copenhagen, Denmark. Zero-baseline, C1 Zero-baseline, L Short-baseline, C1 Short-baseline, L1 Receiver Session Mean [mm] North East Up Std. dev. [mm] North East Up LEA-6T Zero-baseline Short-baseline Redidual [cm] 1 Redidual [m] Table 1. Positioning statistics for the zero-baseline and short-baseline in North, East and Up local topocentric coordinate system. 1 Redidual [cm] Redidual [m] Table 2. Statistics values of phase residuals for u-blox LEA-6T zero-baseline Figure. Code and phase residuals plotted with respect to elevation angle for zero- and short-baseline. Note the different units on the vertical axes. PRN Tab. 1 displays statistics (mean and standard deviation) to get an impression of the data quality. The mean for short-baseline is excluded from the table as the true values are not known. Both test cases indicate that millimeter to centimeter level positioning precision can be reached. However, a significant difference on the noise level between the horizontal and vertical components in both test cases. In addition, the standard deviations for the short-baseline case are larger than those for the zerobaseline case Residuals analysis As mentioned, the data processing was done in a kinematic mode, epoch-by-epoch. According to the mathematical model, the residuals of the observations are expected to have zero mean. The residuals give an impression on the observation noise and indicate biases and anomalies if any. One should keep in mind, that in general, filtering and smoothing reduces the noise level at the expense of increasing observation time-correlation. In addition, multipath presence in the short-baseline may introduce time-correlation to the observations. Elevation range Statistics [cm] Min Max Std Samples Cycle slips Table 3. Statistics values of phase residuals for u-blox LEA-6T short-baseline. PRN Fig. shows the code and phase residuals plotted with respect to the elevation angle to the satellite in the zero- and short-baseline test. A simple comparison shows that the residuals (both C1 and L1) for low elevation satellites are generally noisier than those of high elevation satellites. Tab. 2 and 3 present statistical values (minimum, maximum and standard deviation) of the phase residuals over the full 2h 3min observation period, for zero-baseline and short-baseline, respectively. The peak-to-peak variations for the phase residuals are less than ±1cm (i.e.,. cycles for L1 frequency), which agrees with the half-cycle ambiguity resolution mentioned in 2.2. The standard deviation is less than 2. cm for all satellites in view. 4 Elevation range Statistics [cm] Min Max Std Samples Cycle slips

5 31 August - 2 September 211, Copenhagen, Denmark. Table 4. Positioning statistics of u-blox LEA-6T receiver when connected to different antennas. Receiver Antenna Mean [mm] RMS [mm] Samples North East Up North East Up u-blox Leica Ashtech u-blox Static positioning with geodetic antennas Two other trial tests were conducted using geodeticquality antenna connected to the LEA-6T receiver. In order to evaluate the positioning performance of such configuration, another dual-frequency receiver from the same manufacturers as the antenna was connected to the same antenna via a signal splitter. The first trial test setup consisted of LEIAX122 antenna, Leica GPS12 and LEA-6T receivers. The test was deployed on Jun 8th, 211. The second trial involved an ASH7228D antenna, Ashtech Z-XII and LEA-6T receivers. This testing was conducted on Jun 9th, 211. Both trial tests covered the same time interval, i.e. 9:3: - 12:: (GP- STime). The observations were collected at a 1-second interval and with an elevation cut-off mask of degrees. In addition, a commercial network RTK reference service was used to generate VRS (virtual reference station) data and determine the benchmark coordinates. The dual-frequency positioning solutions for each observation epoch were considered as reference when evaluating the accuracy of LEA-6T positioning coordinates. The results show cm-level accuracy in all three components. The horizontal components indicate few millimeters biases and a RMS (root-mean-square) factor around 1 cm. On the other hand, the vertical component looks like having a few centimeter biases. However, these biases contain in fact the vertical offset of the phase center of the antenna used in the corresponding test: 66.7 mm (LEIAX122 antenna) and 6.4 mm (ASH7228D) antenna[7]. Tab. 4 presents the positioning statistics (mean and RMS) obtained when LEA-6T receiver is used in combination with different antenna types. The statistics are calculated excluding the convergence period. The last column gives the number of samples used in the statistics calculation. For the Leica trial test, the reference solution was not valid due to the lack of reference data for the first 28 minutes. On the other hand, for the Ashtech trial test, the number of samples is smaller due to a slip in the Ashtech receiver setup settings, which collected data only at a 1- second interval. As a result, the reference coordinates were available only for one tenth of u-blox positioning solutions. Figure 6. u-blox static positioning with different antennas: Ashtech (a), u-blox (b) and Leica (c). 3.. Kinematic real-time positioning The previous reported results are obtained using RTKLIB post-processing baseline analysis application (i.e., RTK- POST). Although dedicated for post-processing, the application is run in kinematic mode meaning solution estimation on an epoch-to-epoch basis. Moreover, the positioning solutions are obtained using only forward direction. This allows us to simulate kinematic real-time testing. On the other hand, RTKIB library also includes an application dedicated to real-time positioning, namely RTKNAVI. The RTKNAVI application program can be run on a PC, uses receiver s carrier-phase measurement in order to enable ambiguity resolution in real-time computes positioning solutions in real-time and outputs them via the output streams. One two-round pedestrian walking test was conducted to test the above mentioned capability. A VRS station was generated in the proximity of the testing area in order to process the navigation-grade receiver observation. Fig. 7 shows the ground plot of the reference trajectory during the pedestrian walk. The horizontal positioning errors are grouped according to different threshold values: larger than cm (red), between and 2 cm (blue), between 2 and 1 cm (green), and smaller than 1 cm (black). The colors are intended for better visualization of the variation and degradation in accuracy. It can be seen that the precision improves as the time elapses. The second round is shifted with -1 m in North and East di-

6 31 August - 2 September 211, Copenhagen, Denmark. North [m] Round 2 Round 1 Kinematic real-time positioning (pedestrian walking) >.m.2m to.m.1m to.2m <.1m Table. Positioning error distribution with respect to different threshold. Threshold value Processed epochs Number Percentage larger than.m between.m and.2m between.2m and.1m smaller than.1m TOTAL East [m] Figure 7. Ground trajectory for the two-round pedestrian walking test. For illustration purposes, the 2nd round is shifted in North and East direction with -1 and -1 m, respectively. ments using officially documented proprietary format. Several static and kinematic tests are conducted to analyze the precision and accuracy of different positioning techniques. The results demonstrate that low-cost precise positioning at centimeter level is possible with LEA-6T. In a multipath free environment, the accuracy level is better than 2 cm. However, the accuracy decreases in challenging environments and, thus, more robust algorithms should then be applied. In addition, the investigation indicates high positioning solution repeatability even if some systematic errors may still contaminate the data. This level of performance together with the economically affordable devices utilized in this paper support the current trend towards low-cost solutions for accurate or precise positioning determination. As a result, new market areas may be opening for various applications, such as land, marine or aerial surveying, GIS applications, structural monitoring, or early warning systems. ACKNOWLEDGMENTS Figure 8. General view of the surrounding environment for the kinematic pedestrian walking test. rection, respectively. The positioning results indicate a degradation in precision due to the multipath effects. The multipath is generated by the presence of the leaf foliage in the northern part. On the other hand, when there are no restrictions on signal reception (i.e, free sky-view), such as in the eastern part, the precision of the positioning solution is better than 1 cm. Fig. 8 gives a general illustration of the surrounding environment in which the test was conducted. Tab. summarizes the positioning results. The precision of the positioning solution is better than 2 cm for 8.9% of the processed epochs. 4. CONCLUSIONS This paper offers a preliminary analysis on the positioning performance of the high-sensitivity navigation grade u-blox LEA-6T receiver. LEA-6T is the latest u-blox receiver capable to output code and phase raw measure- The first author would kindly acknowledge Mr. Seppo Tötterström, from Geotrim Oy for providing access to VRS data. Mr. Panu Salo, Department of Surveying, is thanked for support in the field operations. REFERENCES [1] Alkan, R. M. (21). Development of a Low-cost Positioning System Using OEM GPS Receivers and Usability in Surveying Applications. In Proceedings of the XXIV FIG International Congress 21, Sydney, Australia. [2] Estey, L. H. and Meertens, C. M. (1999). TEQC: The Multi-Purpose Toolkit for GPS/GLONASS Data. GPS Solutions, 3(1): [3] Fastrax (27). itrax3 OEM GPS Receiver Module. Fastrax Ltd., itrax3 datasheet. [4] Hill, C. J., Moore, T., and Dumville, M. (21). Carrier Phase Surveying with Garmin Handheld GPS Receivers. Survey Review, 36(282): [] Masella, E., Gonthier, M., and Dumaine, M. (1997). The RT-STAR: Features and Performance of a Low- Cost RTK OEM Sensor. In Proceedings of the 1th 6

7 31 August - 2 September 211, Copenhagen, Denmark. International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GPS 1997), pages 3 9, Kansas City, MO. [6] Neil D. Weston and Volker Schwieger (21). Cost Effective GNSS Positioning Techniques. FIG Publication No 49, FIG Commission Publication. Copenhagen, Denmark. [7] NGS (211). GPS Antenna Calibration. Online. Available at: gov/antcal/images/ant_info.abs. [8] Rizos, C., Han, S., and Han, X. (1998). The RT- STAR: Features and Performance of a Low-Cost RTK OEM Sensor. In Proceedings of the 11th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GPS 1998), pages , Nashville, TN. [9] RTKLIB (211a). Manual version On-line. June 11, 211. [1] RTKLIB (211b). RTKLIB: An Open Source Program Package for GNSS Positioning. On-line. http: // [11] Schwieger, V. (23). Using Handheld GPS Receivers for Precise Positioning. In Proceedings of the 2nd FIG Regional Conference, Marrakech, Morocco. [12] Söderholm, S. (2). GPS L1 Carrier Phase Double Difference Solution Using Low Cost Receivers. In Proceedings of the 18th International Technical Meeting of the Satellite Division of The Institute of Navigation (ION GNSS 2), pages , Long Beach, CA. [13] Takasu, T. and Yasuda, A. (28). Evaluation of RTK-GPS Performance with Low-cost Singlefrequency GPS Receivers. In International Symposium on GPS/GNSS 28, Tokyo, Japan. [14] Takasu, T. and Yasuda, A. (29). Development of the low-cost RTK-GPS receiver with an open source program package RTKLIB. In International Symposium on GPS/GNSS, Jeju, Korea. [1] Teunissen, P. J. (199). The least-squares ambiguity decorrelation adjustment: a method for fast GPS integer ambiguity estimation. Journal of Geodesy, 7(1-2):6 82. DOI: 1.17/BF [16] ublox (21). u-blox 6 GPS Modules. ublox ag., LEA-6 - Data Sheet. Document no. GPS.G6-HW- 94-C. [17] UNAVCO (21). TEQC The Toolkit for GPS/GLONASS/Galileo/SBAS Data. On-line. Available at: software/teqc/. [18] UNAVCO (211). Permanent GNSS/GPS Stations - Budgets. On-line. Available at: categories/gnss+permanent+stations/. 7