The Integration of GPS and Pseudolites for Bridge Monitoring

Transcription

1 The Integration of GPS and Pseudolites for Bridge Monitoring J. Barnes, C. Rizos, H-K. Lee Satellite Navigation and Positioning Group, School of Surveying and Spatial Information Systems University of New South Wales, Sydney, NSW 252, Australia G. W. Roberts, X. Meng, E. Cosser, A. H. Dodson Institute of Engineering Surveying and Space Geodesy, The University of Nottingham, Nottingham, UK Abstract. One application in structural engineering is measuring the movement of suspension bridges. Under unfavourable wind conditions or heavy traffic loads cable suspension bridges may move up to a few metres. Therefore, a positioning system used to monitor bridge movements can provide extremely valuable information. For such monitoring applications it is desirable for the positioning system to deliver equal precision in all position components, all the time. GPS has the potential to deliver cm-level positioning accuracy, through use of the very precise carrier-phase measurement. However, the same level of accuracy cannot be guaranteed in all three-position components, twenty-four hours a day, in every situation. Additionally, GPS signals are very weak and easily obstructed, especially in built-up urban environments. One solution to improving the GPS satellite geometry and the availability of ranging signals is to use ground-based transmitters of GPSlike signals (called pseudolites ). A major advantage of pseudolite devices is the ability to control where they are located, improving a geometrically poor GPS constellation. In this paper proof of concept for the use of pseudolites in the monitoring of suspension bridges is demonstrated through an experimental trial conducted on the Parsley Bay suspension footbridge in Sydney, Australia. Keywords. Pseudolite, GPS augmentation, structural deformation 1 Introduction The Global Positioning System (GPS) has been used for many years for the deformation monitoring of man-made structures such as bridges, dams and buildings, as well as geodetic applications, including the measurement of crustal motion, and the monitoring of ground subsidence and volcanic activity. One application in structural engineering is measuring the movement of suspension bridges. Under unfavourable wind conditions or heavy traffic loads cable suspension bridges may move up to a few metres. If bridge movements can be precisely measured then bridge designers can compare real and predicted deformation. With continuous or periodic measuring of a bridge s movement, the structural integrity can be monitored and warn of potential failure. For the monitoring of man-made structures, and specifically suspension bridges, it is desirable for the measurement system to deliver equal precision in all position components, all of the time. GPS has the potential to deliver cm-level positioning accuracy, through use of the very precise carrier-phase measurement. However, the accuracy, availability, reliability and integrity of the GPS position solutions are very dependent on the number and geometric distribution of the available satellites. Therefore, the positioning precision is not the same in all three components, and large variations in positioning precision can be expected during a 24-hour period. This situation becomes worse when the line-of-sight to GPS satellites becomes obstructed, such as in urban environments. Also, GPS data below an elevation of approximately 15 degrees is usually of poorer quality and typically not used, due to difficulties associated with error modelling and susceptibility to being obstructed. This fact and the geometric distribution of the GPS satellite constellation results in 2 to 3 times worse accuracy in the height component, in comparison to horizontal components. Additionally, in mid and high latitude areas (>45 degrees) the accuracy in the North component is worse than the East component, due to the 55 degree inclination of GPS satellites (Meng et al., 22).

2 One solution to improving the GPS satellite geometry and the availability of ranging signals is to use ground-based transmitters of GPS-like signals (called pseudolites ). A major advantage of pseudolite (PL) devices is the ability to control where they are located, improving a geometrically poor GPS constellation. With four or more pseudolite devices, positioning independent of GPS is possible, with the advantage of complete control over the fixed ground based signal transmission. The Satellite Navigation and Positioning (SNAP) group at The University of New South Wales (UNSW) has been actively conducting research into high precision combined kinematic GPS-PL for the past three years. This has led to the development of innovative carrier-phase single-epoch software that can process both GPS and PL data. Due to the comparatively small separation between PLs and user receivers, there are some challenging modelling issues, such as non-linearity, PL location errors, tropospheric delays, multipath and noise. These issues have been discussed in a series of papers by SNAP researchers (Barnes et al., 22a; Wang et al., 21; Wang et al., 2; Dai et al., 2). In addition, not all GPS receivers can track PL signals, and there are near-far signal strength issues. Because of these difficulties PLs are not yet a mainstream off-the-shelf technology. Deformation monitoring is one application where SNAP has explored the use of PLs (Barnes et al., 22b; Dai et al., 2). More recently the SNAP group, together with the IESSG at Nottingham University (UK), have conducted research into the use of PLs for the monitoring of suspension bridges. The IESSG have been interested in the monitoring of suspension bridges using GPS and accelerometers for the past few years, and their research is documented in such papers as Meng (22), Meng et al. (22), Dodson et al. (21). In this paper the results and analyses are presented of the augmentation of GPS with PLs for a trial conducted on a suspension footbridge in Sydney, Australia. Fig. 1 Parsley Bay suspension footbridge trial. constructed from wood, whilst the cabling and sides are steel. The bridge, now classified by the National Trust, is approximately 55 metres long and 2 metres wide. On 16 January 23 a trial was carried out on Parsley Bay Bridge to test the suitability of a PLaugmented GPS system for bridge monitoring. Two IntegriNautics PLs (PL12 and PL32) were used in the experiment, and their configuration is illustrated in Figure 2. Their locations were selected so that distances to both the reference receiver antenna (base) and the antenna mounted on the bridge (rover) were approximately equal, and with a clear line-of-sight. This was to ensure that the base and rover GPS receiver could track the PL signal with approximately the same signal strength. Additionally, the locations of the PLs were selected with a relatively unobstructed sky to allow the positions to be determined using GPS. The elevation angles of the PLs from the base antenna were close to zero, and at negative elevation angles from the rover antenna (PL and PL degrees). Leica SR53 GPS receivers were used to survey the locations of the PLs with respect to the base 2 Parsley Bay Suspension Footbridge Trial Parsley Bay is a narrow inlet on the east side of Sydney Harbour, and is a popular picnic and swimming location. In 191 a cable suspension footbridge (Figure 1) was constructed across the bay to improve pedestrian access between the two shores of the bay. The bridge was built to the design of Edwin Sautelle, who was Town Clerk and Engineer of Vaucluse Council, and cost 5 pounds. The bridge towers, deck and handrails are Fig. 2 Location of PLs, base and rover receiver.

3 antenna, by collecting approximately 3 minutes of data at a 1Hz rate. For the tracking of PL and GPS signals during the trial, dual frequency NovAtel Millennium (OEM3) GPS receivers were used. The NovAtel receiver allows the user to request particular PRN codes to be tracked, and this is a basic requirement for a PL-tracking GPS receiver. The PLs used Micropulse passive patch antennas, and were assigned GPS PRN codes 12 and 32. They operated in a pulsed mode with a 1% duty cycle, and their power was adjusted to give good signal-to-noise ratios (above 45-dBHz) at both the base and rover, but not too strong so as to jam the GPS signals. During the test approximately 2 hours and 2 minutes of data at a 2Hz rate were recorded from the NovAtel receivers, using laptop computers via a serial interface. On six occasions during the trial movement of the footbridge was intentionally induced. This was achieved by 4 people rocking the bridge from sideto-side. Five of these occurred approximately between 18-33s and one at 76s, from the start of the trial. 2.1 Geometry Analysis of GPS-only and PL Augmented GPS (GPS-PL) It is important to investigate the rover geometry for GPS-only and pseudolite augmented GPS (GPS-PL) configurations during the trial period. There were between 4 and 7 satellites available above 15 degrees during the trial. Figure 3 shows the Dilution of Precision (DOP) in East, North and Vertical directions, together with the number of SVs/PLs available for GPS-only and GPS-PL configurations. It is clear from Figure 3 that when the number of GPS satellites drops to 4, just after 7 seconds, the satellite geometry is extremely poor, with all DOP values greater than 8! DOP values as large as this are likely to result in very poor positioning precision. It should be noted that the drop in satellites to 4 (above 15 degrees elevation) is purely due to a weak satellite constellation and not an obstructed sky-view. Also, a drop in satellite numbers from 6 to 5 has a major impact on the geometry, indicated by spikes in the DOP time series. From Figure 3 it is clear that the pseudolites improve the GPS-only geometry greatly. When the number of GPS satellites drops to 4 and 5 in the GPS-PL configuration, the DOP values are not significantly affected, with all DOP values below 2.4. To assess the improvement in the GPS constellation through augmentation with PLs the Number of SVs & PLs and DOPS NSVs VDOP 2 NDOP EDOP Number of SVs & PLs and DOPS VDOP NDOP NSVs and NPLs mean, maximum and minimum DOP values were computed for both configurations and are given in Table 1. For this analysis, epochs with only 4 satellites were not used, since epochs with such high DOP values would not result in cm-level positioning and be rejected by most GPS processing software. Overall, for both GPS-only and GPS-PL configurations, the best geometry is in East, then North, and lastly Vertical. For the GPS-PL configuration the mean DOP values are all less than 1. The improvement in the geometry (% Diff. in Table 1) when using PLs is greatest in the Vertical (42.6%), then North (2.9%) and least in East (17.2%). Table 1. Summary of DOP values for GPS-only and GPS augmented with pseudolites, computed where there are more than 4 SVs. GPS-only GPS-PL % Diff. Max EDOP Min.6.56 Mean Max NDOP Min Mean Max VDOP Min Mean >8 EDOP Fig 3. DOPs and NSVs for GPS-only (top) and GPS-PL (bottom) configurations.

4 2.2 GPS and Pseudolite Data Processing The baseline software developed by the SNAP group was used to process the GPS and PL data. For all the data processing a satellite elevation mask of 15 degrees was used, and the final solution type was L1 double-difference ambiguity-fixed. For PLs used in a static environment, over short distances, multipath is the dominant error and theoretically is a constant. For this bridge trial it was expected that the movements would be relatively small (less than one metre), and a constant pseudolite multipath bias was assumed. The multipath bias associated with the pseudolite must therefore be determined before using the PL carrier-phase data in the position computations. The methodology for this is described below. Using only GPS data, an On-The-Fly (OTF) kinematic solution was first computed. The average rover position of the bridge was then computed using solutions with more than 4 satellites (due to the very poor DOP as indicated in section 2.1). Then, using the computed rover coordinates, carrier phase double-difference residuals were calculated Double difference residual (metres) (L1).1 mean 3. mm Epoch stdev 8.7 mm 14 12(L1).5 Double difference residual (metres) mean 53.1 mm Epoch stdev 13.1 mm 14 32(L1).5 Double difference residual (metres) mean 7.8 mm Epoch stdev 12.3 mm Fig. 4. Double difference residuals for SV3 (top), PL12 (middle) and PL32 (bottom). for the GPS and PL data using SV14 as a reference satellite (assumed to have no long term biases). The double-difference residuals for the satellites were as expected, with small mean values (less than 3.5mm), indicating no significant biases. A typical satellite double-difference residual time series is given in Figure 4, for SV3 (elevation degrees). The double-difference residual time series for the two pseudolites (in Figures 4) have significant constant biases, attributeable to multipath. The mean values of the time series, in millimetres, are 53.1 (PL12) and 7.8 (PL32). There are some time series fluctuations, in the GPS and GPS-PL time series, which are due to bridge movement and/or multipath. Using the mean values, the PL data was corrected for the constant bias (due to multipath). Finally, in the data processing an OTF GPS-PL solution was computed. Where it was not possible to resolve ambiguities OTF, the average position of the bridge was used to resolve the carrier-phase ambiguities. 2.3 Results of GPS-only and GPS-PL Positioning Figure 5 shows the GPS-only positioning results in East, North and Vertical (mean-centred). From approximately epoch 7s onwards, the position time series shows large position variation. This variation occurs when the number of GPS satellites drops to 4, resulting in very poor GPS geometry (see section 2.1). As mentioned in section 2, bridge movement was induced on six occasions, and one of which occured around epoch 76s. However, it is difficult visually to separate the actual bridge movement from position variation induced by poor geometry. For the other 5 instances of induced movement (18-33s) the East and North time series show variation of several centimetres, and little variation in height. Figure 6 shows the GPS-PL positioning results for East, North and Vertical. Compared with the GPS-only positioning result (Figure 5), there is not a large variation in the position time series from epoch 7 s onwards. However, visually there is variation in the North position time series of up to 3cm around epoch 76s (due to induced movement). Like the GPS-only solution, the induced bridge movement, around epochs 18-33s, is as large as 6.5cm and 1.cm in the East and North components respectively. Visually, it is also very clear that the GPS-PL position solution has less noise than the GPS-only solution. To demonstrate this fact standard deviations of the East, North and Vertical position

5 East stdev mm East stdev 6.9 mm North stdev mm.1.1 North stdev 1.3 mm Up stdev mm.1.1 Up stdev 6. mm Fig. 5. GPS-only positioning result in East (top), North (middle) and Vertical (bottom). Fig 6. GPS-PL positioning result in East (top), North (middle) and Vertical (bottom). time series are given in Table 2. These standard deviations are computed using data with more than 4 satellites, and not in periods of induced bridge movement. When GPS is augmented with PLs, the greatest improvement in the standard deviations is in the Vertical (58.7 %), then North (28.%) and lastly East (27.1 %). The level of improvement is slightly better than that expected through geometry analysis (section 2.1). For the GPS-only solution the standard deviation in the Vertical (13.1mm) is approximately twice that of the horizontal components. Whereas for the GPS-PL solution the Table 2. Standard deviations of East, North and Vertical time series for GPS-only and GPS-PL, computed when there are greater than 4 SVs and no periods of induced bridge movment. L1 single epoch Standard deviation (mm) solution type East North Vertical GPS-only GPS-PL % Diff position component standard deviations are all less than 5.5mm, the Vertical precision is similar to the North component, and the East is slightly better. This result is the same as that expected through the geometry analysis referred to in section Analysis of GPS-PL result Does the GPS-PL solution time series correctly depict the actual movements of the bridge? For this it is useful to consider the orientation of the bridge with respect to the position components. From Figure 7 it can be seen that the centreline of the bridge runs approximately in a NE-SW direction. Therefore from the induced side-to-side bridge movement deviations in both horizontal components are expected (slightly more in North), and little in the Vertical. From Figure 6 this situation is clearly the case. Using the data period from approximately 28-33s, the maximum deviations in the positioning time series are 6.5, 1. and 2.6cm for East, North and Vertical respectively. For the Vertical component any movement is visually difficult to see above the

6 noise of the solution and the maximum horizontal deviation is 11.6cm. Overall, the variations in the GPS-PL position time series are of the orders of magnitude expected. Acknowledgments The authors would like to thank Woollahra Municipal Council for allowing the trial to be conducted at the Parsley Bay Bridge. The collaborative research between SNAP and IESSG was aided through an Australian Research Council (ARC) International Linkage grant. References Barnes, J., J. Wang, C. Rizos, T. Nunan, C. Reid (22a). The development of a GPS/pseudolite positioning system for vehicle tracking at BHP Billiton steelworks. In: Proc of 15th International Technical Meeting of the Satellite Division of the US Inststitute of Navigation (ION-GPS 22), September, Portland, Oregan, pp Fig. 7. Orientation of parsley bay bridge and relative locations of PLs and GPS receivers. 3. Concluding remarks This paper has shown that there are situations when GPS cannot provide cm-level positioning accuracy twenty-four hours a day when using a 15 degree elevation mask, primarily due to poor constellation geometry. By augmenting GPS with pseudolites it has been demonstrated that very poor geometry situations can be avoided. The introduction of the pseudolites in this trial improved the positioning precision by the greatest amount in the Vertical (58.7%), and by similar amounts for East (27.1%) and North (28%), resulting in similar positioning precision (sub-cm) for both horizontal and vertical components. The pseudolite measurements suffered from constant biases due to multipath error, of which the largest was nearly 7.1cm. However the (almost) static nature of the trial allowed the multipath bias to be calibrated and removed from the pseudolite measurements. All the induced bridge movements during the trial were visually apparent in the position time series and resulted in a maximum horizontal deviation of nearly 12cm. The vertical bridge movements were too small to be detected visually. For a continuous deformation monitoring system there are situations when the GPS geometry can be very poor, and augmentation using pseudolites is one solution to the problem. Barnes, J., J. Wang, C. Rizos, T. Tsujii (22b). The performance of a pseudolite-based positioning system for deformation monitoring. In: Proc of 2nd Symposium on Geodesy for Geotechnical & Structural Applications, May, Berlin, Germany, pp Dai, L., J. Zhang, C. Rizos, S. Han, J. Wang (2). GPS and pseudolite integration for deformation monitoring applications. In: Proc of 13th International Technical Meeting of the Satellite Division of the US Institute of Navigation (ION-GPS 2), September, Salt Lake City, Utah, pp Dodson, A.H., X. Meng, G. Roberts (21). Adaptive method for multipath mitigation and its applications for structural deflection monitoring. In: Proc of International Symposium on Kinematic Systems in Geodesy, Geomatics and Navigation (KIS21), 5-8 June 21, Banff, Alberta, Canada, pp Meng, X. (22). Real-Time Deformation Monitoring of Bridges Using GPS/Accelerometers. PhD thesis, The University of Nottingham, Nottingham, UK, 239 pp. Meng, X., G.W. Roberts, A.H. Dodson, E. Cosser, C. Noakes (22). Simulation of the effects of introducing pseudolite data into bridge deflection monitoring data. In: Proc of 2nd Symposium on Geodesy for Geotechnical and Structural Engineering, May 22, Berlin, pp Wang, J., T. Tsujii, C. Rizos, L. Dai, M. Moore (21). GPS and pseudo-satellites integration for precise positioning. Geomatics Research Australasia, 74, pp Wang, J., T. Tsujii, C. Rizos, L. Dai, M. Moore (2). Integrating GPS and pseudolite signals for position and attitude determination: Theoretical analysis and experiment results. In: Proc 13th International Technical Meeting of the Satellite Division of the US Institute of Navigation, September, Salt Lake City, Utah, pp