CC BY
21
УДК 528.2:629.78 Joël van Cranenbroeck
Leica Geosystems AG, Switzerland (Leica Geosystems)
MONITORING APPLICATIONS USING LEICA SINGLE TO MULTI FREQUENCY GNSS SIGNAL PROCESSING SOLUTION
Key words: GNSS, Single and Dual Frequency, Continuous Monitoring.
ABSTRACT
In many monitoring applications, such as tall buildings, bridges and volcanoes, GNSS offers significant advantages over other measurement techniques. GNSS allows a high rate of measurement and long distances between the control and monitoring points and does not require line of sight to the control points.
The traditional dual-frequency GNSS receivers used in surveying are high accuracy but also relatively high cost per monitored point and therefore often prohibitive for the deployment of a GNSS monitoring network.
Until recently, cost effective L1 sensors and software have not been able to provide the necessary level of accuracy and reliability. A new solution from Leica Geosystems provides real time and post processed RTK positioning with L1 only GNSS receivers for monitoring applications.
The solution is built on Leica's RTK positioning algorithms, which have proven world-class performance in the surveying industry. The positioning algorithm used in the Leica RTK GNSS receivers has been implemented in the Leica Spider reference station software and tuned for monitoring applications, giving Leica Spider the capability to compute real time ambiguity fixed solutions for single and dual frequency GNSS in addition to its powerful site configuration and data management tools.
A direct link has been made between Leica Spider and Leica GeoMoS, Leica's geodetic monitoring software, so that users can combine GNSS with the sophisticated terrestrial measurement capabilities of Leica's robotic total stations and utilize GeoMoS's flexible messaging and data analysis capabilities. In addition the RINEX data logged by GNSS Spider may be automatically post processed for users with the highest accuracy and reliability requirements.
This paper presents results from the system, including using a new ambiguity resolution technique, called quasi-static initialisation, designed for single frequency monitoring. Results from L1 processing are compared to a dual frequency solution in terms of accuracy and reliability. Data was collected with a range of baseline lengths up to 20km in medium multipath environments, typical of many monitoring applications. The L1 system is shown to have remarkable accuracy and reliability, especially in terms of price versus performance.
1. INTRODUCTION
GNSS is a very interesting tool for monitoring because it has a number of distinct advantages over terrestrial positioning technologies. GNSS is able to measure at high rates with low latency, operate in all weather conditions, has synchronized measurement, does not require line of site to ground marks/targets, can measure over long baselines, has low maintenance and a long service life and can provide timing for other sensors, such as accelerometers.
These unique characteristics make GNSS particularly interesting for monitoring large structures such as high-rise buildings and suspension bridges, but also for seismic and land slide applications (e.g. van Cranenbroeck and Troyer, 2004) and for the provision of control for other instruments, such as robotic total stations, in unstable areas.
The main disadvantage of GNSS for monitoring is cost. Each point to be measured must have an antenna, a receiver, ground mark, power, communications and, possibly, protection against lightning and vandalism or theft. Hence, lower cost single frequency (L1 only) GNSS receivers are attractive.
Single frequency GNSS receivers do not require as many tracking channels making them cheaper and more energy efficient and also do not require proprietary algorithms to extract high-quality measurements from the encrypted code on the L2 frequency. The disadvantage of single frequency receivers is that much less measurement data is available to help resolve the carrier phase ambiguities. Also, many lower cost single frequency receivers have poor multipath mitigation capabilities and are more prone to having cycle slips than their higher cost dual frequency cousins.
In monitoring applications, accuracy if of paramount important, so only ambiguity-fixed positions are of interest. A highly reliable ambiguity resolution strategy is needed to prevent wrong fixes, which will be detected immediately by the monitoring system as an apparent movement.
In this paper a comparison is made between the performance of a single frequency and a dual frequency monitoring system in terms of accuracy and reliability. The processing kernel that has been developed for this testing is based on that used in Leica Geosystems high-end RTK GNSS sensors and the LGO software.
The kernel, which is integrated into the Leica GNSS Spider reference station and GNSS monitoring software, is able to process single and dual frequency data in real time and post processing. Two ambiguity resolution techniques are used for this testing: kinematic on the fly (OTF) and a quasi-static approach. The OTF technique allows full dynamics of the rover antenna suitable for use in formula one racing. The quasi-static approach uses assumes lower dynamics such as would be experienced in most monitoring applications.
An overview of the GNSS Spider and related software and hardware is given. The test setup is described and empirical results are presented that show the comparative performance of single and dual frequency monitoring.
2. ADVANCED GNSS MONITORING
2.1 Overview
The traditional approach to real time GNSS monitoring is to deploy RTK enabled receivers to the field, which as fed corrections from a nearby reference station. This distributed processing approach has some distinct disadvantages:
- Two communications lines are required per point that is measured (one to receive the corrections and one to transmit the resulting coordinates),
- Only one baseline can be computed per point,
- Single frequency RTK is not supported,
- Post processing is not possible, and
- Archiving of the raw observations is not possible.
In the decentralized approach used by Leica GNSS Spider, only a single communication channel is required to send the raw observations to the monitoring server. Multiple baselines may be computed for each point using different reference stations or processing parameters. Single frequency RTK is supported, as is post processing and archiving of both raw data and results. In the case of unreliable communications, it is also possible to log directly in the memory of the GNSS and then download the data periodically for post processing, rather than relying on having a permanent open communications channel.
The Leica GNSS Spider software is a dual-purpose software. It offers comprehensive GNSS reference station capabilities for the configuration and control of GNSS sensors, archiving of data and dissemination of correction data for single-base and network RTK positioning.
Leica GNSS Spider boasts a state of the art network-processing kernel designed for the new Master-Auxiliary concept (Leica Geosystems, 2005). In addition to the reference station capabilities, GNSS Spider has advanced GNSS baseline processing capabilities for monitoring applications. The marriage of reference station and GNSS monitoring features produces a flexible and powerful application with sophisticated communications, processing, data management and security functionality. GNSS Spider may be combined with the Leica GeoMoS geodetic monitoring software for integration with robotic total stations, inclination and other geotechnical sensors and to leverage its advanced limit checks, messaging and analysis features. The baseline processing in GNSS Spider is divided in two parts: real time processing and post processing.
2.2 Real Time Monitoring With GNSS Spider
The real time processing kernel is based on that used in the GX1230 RTK rover, but has been modified for monitoring applications. The Leica SmartCheck technology, which is an evolution of the repeated search process described by
Euler et al. (2000), is used to continuously re-verify the ambiguity fix to ensure the highest reliability. With this improved kernel GNSS Spider is able to compute RTK-fixed positions from both single and dual frequency data.
Three ambiguity resolution techniques are available: Kinematic on the fly (also known as OTF or While Moving initialisation) , Initialisation on Known Marker (IOKM) and Quasi-Static (QSI). The OTF ambiguity resolution allows for full receiver dynamics during the initialisation at the cost of reliability, especially for single frequency processing. The IOKM ambiguity resolution assumes strictly limited receiver dynamics (which is not practical for monitoring) but has much higher reliability. The QSI technique is combination of the previous two techniques - it allows for the antenna to be in motion during the initialisation but not to the same extent as OTF initialisation.
2.3 Post Processing
The post-processing kernel used in GNSS Spider is based on that used in LGO. Like with the real time processing, a repeated search process is used to ensure highly reliable ambiguity resolution. In addition, the initialisation on float marker presented by Kotthoff et al. (2003) is used to further improve the reliability. Post-processing intervals of between 1 minute and 24 hours are possible for dual frequency data and between 10 minutes and 24 hours for single frequency data.
2.4 Algorithm Validation
In order to validate the QSI technique and to measure the performance of the single frequency baseline processing (both real time and post processing) in GNSS Spider, test data has been collected for a range of baselines from 30m to 20km (shown in Table 1). Position results derived from four days of 1Hz data collected using five Leica dual frequency receivers are presented in this paper.
Each baseline was processed in real time using both L1 only data and using L1/L2 data and with both OTF and QSI ambiguity resolution for a total of four real time solutions per baseline. Each baseline was also post processed using L1 only data and using L1/L2 data with periods of 10 minutes, 30 minutes and 1 hour, also giving six solutions per baseline for the post processing.
Table 1. List of baselines processed
Num Station1 Station2 Baseline Length
1 HEER BRON 30m
2 HEER KEW1 3.3km
3 HEER RUTH 14km
4 HEER FLDK 20km
The antennas for the sites were placed in non-ideal locations with obstructions and high multipath environments to simulate the conditions of real monitoring sites.
Note that in this paper the raw positions calculated from the processing kernels are analyzed. In a real monitoring installation using the Leica GNSS Spider and/or Leica GeoMoS software, it would be possible to smooth and/or average the results for higher precision and accuracy. In the case of post processing, longer intervals could be used to further improve the precision and accuracy.
3 RESULTS
3.1 Real Time Processing
The first value that is analyzed is the availability of ambiguity-fixed positions. Ambiguity-fixed positions provide the highest accuracy possible with GNSS and so are preferred for monitoring applications. Ambiguity resolution is more difficult when processing single frequency data because of the reduced redundancy and limited information on atmospheric error that is available from the system.
Figure 1 shows the percentage of fixed solutions for each baseline, each frequency and each ambiguity resolution technique. The 1Hz data spanning four days gives a total number of 345,600 possible fixes. As might be expected, the dual frequency processing is gives very high availability of ambiguity fixed positions (approximately 99%) for all baselines, even using OTF initialisation. Single frequency processing using OTF initialisation clearly has more difficulty fixing for the longer baselines. The QS initialisation gives consistently high results for both single and dual frequency data and all baseline lengths.
100
1 (30m) 2 (3.3km) 3 (14km) 4 (20km)
□ L1 (OTF) □ L1/L2 (OTF) □ L1 (QS) □ L1/L2 (QS)
Fig. 1 Availability of Ambiguity Fixed Positions (Real Time Processing)
The availability of an ambiguity fix is related to the reliability of the ambiguity fix. A high availability of fixed positions should not be at the cost of
having more incorrect ambiguity fixes, since wrong ambiguity fixes will result in low accuracy position solutions.
In order to quantify the likelihood of the system to produce invalid or low accuracy positions, the difference between each calculated position and the known coordinates of the site have been compared. All positions that deviated by more than 5cm in horizontal or 10cm in vertical from the true coordinate were flagged as low accuracy position fixed. A position outside this tolerance could be caused by an invalid ambiguity fix or by higher than normal ionospheric activity.
Dual frequency processing is clearly highly reliable, even with OTF initialisation. Single frequency processing with OTF initialisation has some reliability issues, but with QS initialisation is comparable to dual frequency, up to about 10 or 15km. For the 20km baseline the single frequency solution with QS initialisation is also showing signs of having reliability issues due to its limited ability to model atmospheric errors. The conclusion is that in terms of ambiguity resolution, real time single frequency monitoring with baselines of up to about 10km is viable using the quasi-static initialisation. What remains then is to test the accuracy of the solution.
Since it is well known that GNSS is more precise that it is accurate, a measure of accuracy has been used for this analysis. The accuracy is calculated as the standard deviation of the calculated positions about the known coordinates. For this calculation only the coordinates that were within the previously mentioned tolerances were used to isolate accuracy of the position solution from the reliability of the ambiguity fix.
Two components, northing and height, are shown respectively in Figures 2 and 3. Northing was used for the horizontal component because it is typically less accurate than easting because of the satellite geometry. For the shorter baselines (30m and 3.3km) both single frequency and dual frequency solutions are accurate to 4mm or better in northing. Interestingly for the 14km baseline single frequency actually gives higher accuracy, though both are in the order of 1cm for the northing component. For the 20km baseline, the dual frequency solution is clearly more accurate. For the height a similar pattern is seen, though interestingly single frequency has the higher accuracy for the 20km baseline.
0.020 0.018 0.016
>
o ro
o
o <
0.014 0.012 0.010 0.008 0.006 0.004 0.002 0.000
1 (30m)
2 (3.3km)
3 (14km)
4 (20km)
□ L1 (OTF) □ L1/L2 (OTF) □ L1 (QS) □ L1/L2 (QS)
Fig. 2 Accuracy of Northing Component in Real Time (1 Sigma)
0.020 0.018 0.016 IT 0.014 | 0.012
§ 0.010 <
£ 0.008 E?
« 0.006 0.004 0.002 0.000
1 (30m)
2 (3.3km)
3 (14km)
4 (20km)
□ L1 (OTF) □ L1/L2 (OTF) □ L1 (QS) □ L1/L2 (QS)
Fig. 3 Accuracy of Height Component in Real Time (1 Sigma)
3.2 Post Processing
For the post processing, similar tests may be performed. Since more data is used in each solution it can be expected that both reliability and accuracy are much higher in post processing than in real time.
For the 10 minute post processing a total of 576 positions was calculated per baseline over the 4 days of data. A total of 192 position solutions were calculated for the 30 minute post processing. Figure 10 shows the availability of fixed positions for the different baselines and L1 or L1/L2 processing. With 10 minutes of data, the availability of ambiguity fixed positions is greater than 85% for single
frequency data. For dual frequency data or with a period of 30 or 60 minutes, the reliability is at approximately 100%.
As with the real time data, the quality of the fixing was tested by checking the calculated positions against a tolerance of 5cm in horizontal and 10cm in vertical. The higher reliability of post processing is clearly supported by the fact that no positions solutions were outside this tolerance.
In terms of accuracy, Figures 5 and 6 show the corresponding accuracy for post processing. As expected, post processing has a higher accuracy than real time processing. Processing 1 hour or move of data gives the best results as this allows sufficient time for estimation of the tropospheric delay.
100
95
90
85
(A C
o
o w
T3
a>
i! 80 a>
TO
■S 75 c a> o
a> 70
Q.
65
60
1 (30m)
2 (3.3km)
3 (14km)
4 (20km)
□ L1 (10min) □ L1/L2 (10min) □ L1 (30min) □ L1/L2 (30min) □ L1 (60min) □ L1/L2 (60min)
Fig. 4 Availability of Ambiguity Fixed Positions (Post Processing)
0.020 -i
0.018
0.016
— 0.014 >
o re
3
o
o <
c
t o
0.012 0.010 0.008 0.006 0.004 0.002 0.000
It
1 (30m)
2 (3.3km)
3 (14km)
4 (20km)
□ L1 (10min) □ L1/L2 (10min) □ L1 (30min) □ L1/L2 (30min) □ L1 (60min) □ L1/L2 (60min)
Fig. 5 Accuracy of Northing Component in Post Processing (1 Sigma)
0.020 0.018 0.016
E 0.014
% 0.012
§ 0.010
~ 0.008 to
® 0.006 0.004 0.002 0.000
1 (30m)
2 (3.3km)
3 (14km)
4 (20km)
□ L1 (10min) □ L1/L2 (10min) □ L1 (30min) □ L1/L2 (30min) □ L1 (60min) □ L1/L2 (60min)
Fig. 6 Accuracy of Height Component in Post Processing (1 Sigma)
4. CONCLUSIONS
This paper has compared the single and dual frequency GNSS monitoring solutions available in the Leica GNSS Spider software.
The quasi-static ambiguity resolution technique was shown to enable very high availability and reliability RTK positioning with L1 only data. In fact, for baselines up to approximately 10 kilometers the single frequency results using the quasi-static approach were shown to be one the same level of reliability and accuracy as dual frequency.
Post processing of the data was shown to give even higher levels of reliability and accuracy.
REFERENCE
Euler, H.-J. and Ziegler, C., (2000). Advances in Ambiguity Resolution for Surveying Type Applications. In: Proc of ION GPS 2000, Salt Lake City, Utah, September 19-22, 2000.
Brown N., Troyer L., Zelzer, van Cranenbroeck J. (2005). Advances RTK and Post Processed Monitoring with Single Frequency GPS. In: GNSS Hong Kong 2005, December, 2005.
Kotthoff, H., Hilker, C. and Ziegler, C., (2003). Strategy of Reliable Ambiguity Resolution for Static and Kinematic Applications. In: Proc of ION GPS 2003, Portland, Oregon, September 9-12, 2003.
Leica Geosystems, (2005). White Paper: Take it to the MAX! An introduction to the philosophy and technology behind Leica Geosystems' SpiderNET revolutionary Network RTK software and algorithms. Leica Geosystems AG, Heerbrugg, Switzerland, Nune 2005. www.leica-
geosystems.com/common/shared/downloads/inc/downloader.asp?id=5367
van Cranenbroeck, J. and Troyer, L. (2004). GPS Spider for Deformation Monitoring. In: Proc of 1st FIG International Symposium on Engineering Surveys for Construction Works and Structural Engineering, Lenton and Wortley Hall, The University of Nottingham, United Kingdom, 28 June - 1 July 2004. www.fig.net/nottingham/proc/gs_07_cranenbroeck.pdf
© Joel van Cranenbroeck, 2007
