By Richard B. Langley, Simon Banville, and Peter Steigenberger.
For a brief period, and for a few hours on certain days, signals from the first four orbiting Galileo satellites could be received by state-of-the-art multi-frequency, multi-constellation GNSS receivers. Although not intended for actual positioning tests, the satellites did provide a first opportunity to assess the prototype Galileo signals in the positioning domain. The results obtained bode well for the future operational Galileo constellation.
The launch and successful operation of the two Galileo In-Orbit Validation Element (GIOVE) satellites — GIOVE-A and GIOVE-B — followed by the two Galileo In-Orbit Validation (IOV) satellites — ProtoFlight Model (PFM) and Flight Model 2 (FM2) — were important steps in the development of Europe’s Galileo satellite navigation system.
The GIOVE test-bed satellites were orbited to secure the use of the frequencies allocated by the International Telecommunication Union for the Galileo system; to verify the most critical technologies of the operational Galileo system, such as the on-board atomic clocks and the navigation signal generator; to characterize the novel features of the Galileo signal design, including the verification of user receivers and their resistance to interference and multipath; and to characterize the radiation environment of the medium Earth orbits planned for the operational Galileo constellation. The IOV satellites, of which there will be four with two more to be launched this fall, are prototype operational satellites designed to validate the Galileo concept in both space and on Earth. Once all four IOV satellites are in orbit, it should be feasible to carry out positioning exercises using just Galileo satellite signals. It was not intended for the GIOVE plus two initial IOV satellites to be used for positioning demonstrations. However, it turns out that (before GIOVE-A and GIOVE-B were recently decommissioned) for a few hours on certain days, signals from all four satellites could be received simultaneously by state-of-the-art multi-frequency, multi-constellation GNSS receivers.
Dual-frequency measurements from the GIOVE satellites and triple-frequency measurements from the IOV satellites have been archived by a number of continuously operating receivers including those in the COoperative Network for GIOVE Observation (CONGO) and those contributing to the International GNSS Service’s Multi-GNSS Experiment (M-GEX) observing campaign. Before joining the M-GEX campaign as the receiver at station UNB3 at the University of New Brunswick (UNB) in Fredericton, Canada, a Trimble Navigation NetR9 receiver fed by a Zephyr Geodetic II antenna was continuously tested at UNB for a couple of months and its 30-second-interval measurements were locally archived. These measurements included (in the terminology used by the Receiver Independent Exchange (RINEX) version 3 format): C1X, L1X, and S1X (pseudorange, carrier-phase, and carrier-to-noise-density-ratio measurements for combined data-plus-pilot tracking of the Open Service signal on the E1/L1 carrier frequency (1575.42 MHz)); C5X, L5X, and S5X (the corresponding in-phase and quadrature (I+Q) measurements on the E5a carrier frequency (1176.45 MHz)); C7X, L7X, and S7X (the corresponding I+Q measurements on the E5b carrier frequency (1207.140 MHz)); and C8X, L8X, and S8X (the corresponding I+Q measurements on the effective E5a+E5b carrier frequency (1191.795MHz)).
Although the IOV satellites are in synchronized orbits in the same plane with mean orbital periods of 1.70475 orbits per day, those orbits are not coordinated with those of the GIOVE-A and GIOVE-B satellites, which had mean orbital periods of 1.69434 and 1.70960 orbits per day, respectively. (The orbit of GIOVE-B was recently raised, following decommissioning.) This means that all four satellites will not generally be in view at a ground station at the same time. However, at a given location on certain days, four-satellite visibility did occur for periods up to a few hours. We identified several such days but were hampered in our efforts to obtain positioning solutions due to the testing programs of the satellites.
Our first constraint concerned GIOVE-A. The European Space Agency carried out tests with this satellite for more than six years and decided to decommission the satellite for its purposes on June 30, 2012, and switched off the navigation signals. This narrowed our window of possible four-satellite-visibility days. Secondly, the clocks on the IOV satellites were allowed to drift so that their offsets with respect to GPS System Time could be very large with offset values of tens to hundreds of milliseconds. Some GNSS receivers cannot make usable measurements when presented with such large clock offsets. This behavior further limited our windows of opportunity for four-satellite Galileo positioning. Nevertheless, we found that on May 17, 2012, the receiver at UNB successfully tracked the four satellites with a period of common visibility of two and a half hours. See Figure 1 for the time series of the occurrences of actual measurements made by the receiver. Common visibility extended from 03:04:30 to 05:34:30 GPS Time with the receiver tracking the satellites without any elevation-angle cutoff imposed.
In the remainder of this article, we describe the procedures used to obtain precise positions from the measurements, including the technique used to determine precise orbit and clock data for the Galileo satellites, and the results we obtained.
Generating the Orbits and Clocks
GIOVE and IOV satellite orbit and clock parameters are determined at Technische Universität München with a modified version of the Bernese GPS Software in a two-step procedure based on GPS and Galileo observations of 23 CONGO stations. After a common preprocessing step (detection of outliers and cycle slips), GPS and Galileo observations are treated separately. Station coordinates, tropospheric delay parameters and receiver clocks are obtained from GPS observations only. GPS satellite orbits and clocks as well as Earth rotation parameters from the Center for Orbit Determination in Europe (CODE) are kept fixed in this step. In the second step, the ionosphere-free linear combination of E1 and E5a observations is used to estimate the Galileo-related parameters, namely the satellite orbits and clocks. The station coordinates and the troposphere and receiver clock parameters are fixed to the GPS-derived results of the first step. To account for systematic differences between the GPS and Galileo code signals as well as biases between the different receivers, differential code biases (DCBs) are estimated for all stations but one. Separate biases are set up for the GIOVE and IOV satellites. To strengthen the stability of the orbital arc, five daily solutions are combined into a multi-day solution and consistent Galileo clock parameters are recomputed. Only the middle day of the5-day solution is used for the positioning discussed in this article. Based on internal consistency tests and satellite laser ranging residuals, the accuracy of these orbits is assumed to be on the one-to-two-decimeter level.
The Positioning Technique
A preliminary assessment of the quality of Galileo-only positioning could be achieved using the four satellites simultaneously in view at UNB. The second author’s GNSS positioning software was used to process the UNB data. Applying a 7.5-degree elevation cutoff angle to remove low-elevation-angle measurements resulted in an observation session of 1 hour and 48 minutes. The east or longitude dilution of precision (DOP) component starts out at 0.829 at the beginning of this session, gradually dropping to 0.688, and then rising to 1.285 at the end of the session; while the north or latitude DOP component starts out at 2.683 at the beginning of the session, rising to 4.233 at the end (see Figure 2).
Even though the receiver was tracking signals on E1, E5a, E5b, and E5a+b, carrier-phase and code observations on E1 and E5a were selected to match the satellite-clock datum. Ionosphere-free combinations were formed to eliminate first-order ionospheric effects, while the tropospheric delay was modeled using local measurements of temperature, pressure, and relative humidity provided by UNB’s meteorological station. No residual delay was estimated. Phase center offsets (PCO) and variations (PCV) for the Trimble Zephyr Geodetic II antenna were obtained through anechoic chamber calibrations (see Further Reading). The same satellite PCO as the ones used in the generation of the satellite orbits and clocks were applied, and no satellite PCV were considered. Other error sources required for precise positioning were also modeled such as solid Earth tides, ocean tide loading, and phase wind-up.
Since separate biases were set up in the estimation of the GIOVE and IOV satellite clock estimation, the same approach should be used on the user side. Unfortunately, solving for this additional parameter in the navigation filter is not possible when tracking only four satellites. To overcome this limitation, a GIOVE/IOV offset was estimated using 24 hours of combined GPS-plus-Galileo observations in static mode (one position solution for the whole observation period), and was introduced as an additional correction in the Galileo-only solution. The estimated coordinates from this combined solution were also used as a reference in computing the errors in latitude, longitude, and height presented next.
Results and Discussion
Three solutions were computed to demonstrate the quality of Galileo-only navigation. In the first scenario (see Figure 3), ionosphere-free code observations solely contributed to the epoch-by-epoch estimation of receiver position and clock offset. The estimated coordinates are largely contaminated by code noise, which is amplified by a factor of approximately three when forming the ionosphere-free combination. In the absence of redundancy, any errors in the observations (such as noise) propagate directly into the estimated quantities and, in this case, affected particularly the latitude component. An analysis of the noise and multipath characteristics of each signal revealed the presence of time-varying effects in the C5X observations. Further investigations are required to properly identify the cause of those effects. As a result, the root-mean-square (RMS) error of the latitude, longitude and height components were 3.084, 0.658 and 1.617 meters, respectively (see Table 1).
As a second step, both code and carrier-phase observations were combined into a single adjustment (see Figure 4), yielding what is often referred to as precise point positioning (PPP). To accommodate the initial carrier-phase ambiguities, additional parameters were estimated in the filter. While adding carrier phases clearly reduces the noise in the solution, the estimated coordinates do not converge to cm-level accuracies, as typically expected in PPP. Despite weak geometry and range errors, the main reason for poor convergence is again the presence of biases in code observations. Without redundancy, carrier-phase observations only act as a filter for code observations, without reducing the contribution of biases. The RMS errors are 0.422 meters in latitude, 0.150 meters in longitude, and 0.389 meters in height.
To get an independent assessment of carrier-phase observations, a phase-only solution was computed (see Figure 5). For this test, a different methodology was adopted in which we simulated starting the positioning at a known precise location. At the first epoch, the receiver coordinates were constrained to the estimated values from the 24-hour GPS-plus-Galileo positioning solution, and the receiver clock offset was fixed to an arbitrary value (in this case zero). This initial epoch thus allowed estimation of the carrier-phase ambiguities, which remained constant for the rest of the session. For subsequent epochs, the receiver position and clock offset were estimated on an epoch-by-epoch basis. Even though the errors in the initial ambiguity estimates propagated into the following epochs, the estimated coordinates remained at the centimeter level throughout the nearly two-hour common observing period.
We have obtained what we believe to be the first positioning results using observations from the four Galileo satellites launched to date. The results are very respectable given that the observing geometry was far from ideal and there was no redundancy for epoch-by-epoch solutions. Furthermore, the satellites were not operating at a performance level to be expected for the fully operational future constellation. Both GIOVE satellites have been retired and we must now wait for the second set of IOV satellites to be orbited before we can continue our investigations in Galileo-only positioning with live signals.
We thank the operators and station managers of the CONGO network for supplying the data used to model the orbits and clocks of the Galileo satellites.
Richard B. Langley is a professor in the Department of Geodesy and Geomatics Engineering at the University of New Brunswick (UNB) in Fredericton, Canada.
Simon Banville is a Ph.D. candidate in the Department of Geodesy and Geomatics Engineering at UNB. He is also working for Natural Resources Canada on real-time precise point positioning.
Peter Steigenberger is a staff member in the Institut für Astronomische und Physikalische Geodäsie of the Technische Universität München in Munich, Germany.
“Anechoic Chamber Calibrations of Phase Center Variations for New and Existing GNSS Signals and Potential Impacts in IGS Processing” by M. Becker, P. Zeimetz, and E. Schönemann, presented at the IGS Workshop, Newcastle upon Tyne, England, June 28–July 2, 2010. Available online: http://www.igs.org/event/newcastle2010/ (scroll to “0205” and click on “PDF.”)
“A Guide to Using International GNSS Service (IGS) Products” by J. Kouba, an IGS resource.
“Precise Orbit Determination of GIOVE-B Based on the CONGO Network” by P. Steigenberger, U. Hugentobler, O. Montenbruck, and A. Hauschild in Journal of Geodesy, Vol. 85, No. 6, 2011, pp. 357–365, doi: 10.1007/s00190-011-0443-5.