Presentation: “GLONASS and Multi-GNSS in the IGS: Lessons learned from GLONASS Service Disruptions”
G. Beutler, R. Dach — Astronomical Institute, University of Bern O. Montenbruck — DLR, German Space Operations Center U. Hugentobler — Technical University of Munich G. Weber — Bundesamt für Kartographie und Geodäsie, Frankfurt E. Brockmann — Swisstopo, Wabern
13th Meeting of the National Space-Based Positioning, Navigation, and Timing (PNT) Advisory Board June 3-4, 2014, Omni Shoreham Hotel, 2500 Calvert Street NW, Washington DC International Association of Geodesy
Content
- The International GNSS Service (IGS)
- The IGS as a multi-GNSS Service
- The IGS Real time Experiment
- The IGS/MGEX Experiment
- The GLONASS April 1st event: a case study
- The GLONASS April 1st, 2014 viewed by the IGS
- Lessons to be learned
The International GNSS Service (IGS)
- …is the premier source of the highest-quality GNSS data, products, and related standards and conventions
- …is in support of many applications that benefit the scientific community and society
- …is a Service of the International Association of Geodesy (IAG) founded in 1994
- …is operational since more than 20 years
- …is a federation of more than 200 institutions and organizations worldwide
- …is following an open data policy
- …is open to everybody to participate
The IGS in particular operates:
- The MGEX, Multi-GNSS Experiment
- The IGS Real-time Experiment
The GLONASS April 1st Incident: a case study
RINEX status monitoring in Switzerland (swisstopo PNAC network) on 02-Apr-2014 07:35:22 (DOY 092) showed widespread data gaps: many stations showing “data o.k.” (green), “3<5 h missing/last h late” (yellow), and “>5 h missing” (red) across sites including AIGE, AND2, ARDE, BLFT, BOUR, BOU2, BSCN, COMO, DAVO, DAV2, EFFL, ERDE, ETH2, FALE, FHBB, FLDK, FREI, FRIC, ERB3, HABG, HOHT, HOH2, HUTT, JUJO, KALT, KOPS, KRBG, KRLU, LECH, LIND, LOMO, LUZE, MAR2, MARG, MTTI, NEUC, PAYE, PFA2, PRNY, SAAN, SAA2, SAME, SANB, SAR2, SCHA, SIGM, STA2, STOX, STGA, VARE, VISW, WEIN, ZERM, ZIMM, ZIM2.
The GLONASS April 1st Incident in the IGS
Chart of ”# GLONASS satellites tracked by IGS receivers” (from CODE analysis protocol of April 1): Some receivers unaffected, some stopped tracking GLONASS after 9h p.m. (red line), some even stopped GPS tracking. Out of about 180 receivers used by CODE, about 60 were severely affected.
Chronicle of Events (Time UTC)
| Time (UTC) | Chronicle of Events |
|---|---|
| April 1, 21:00 | GLONASS starts transmitting “infected” Broadcast messages simultaneously from all satellites. |
| April 2, 7:00 | All GLONASS satellites transmit again correct broadcast messages (satellite 21 last satellite). |
| April 3, 9:23 | E-Mail by Chris Rizos (IAG President and IGS GB member) to IGS GB et al: “What do you guys know about this? Have you been fielding any questions from media? users?” (wake-up call for IGS) |
| April 3, 10:36 | E-Mail by Tim Springer (ESA Analysis Center) stating in essence that the “normal” IGS processing (final, rapid, ultra-rapid) was not affected (“nobody noticed”) |
| April 4, 14:29 | E-mail by Urs Hugentobler, Chair IGS Governing Board: First IGS-internal analysis of the event. Osculating elements derived from GLONASS Broadcast analyzed. He “sees” rotations of orbital planes, specific to three GLONASS planes |
| Since | Various attempts to explain the signature of the event (from cyber war to more technical and reasonable) |
From CODE Analysis protocol: Number of GLONASS observations reduced by about 10% on April 2nd. → this is why the IGS post-processing products were virtually not affected (orbits are based on one day of observations). Had the incident lasted for > 1 day, the impact would have been more severe.
Satellite position differences (referring to the GLONASS/Broadcast- and CODE/IGS-derived positions [“truth”], in the Earth-fixed equatorial system, between April 1st 20:00 and April 2nd 08:00):
- the differences are of the order of ±200km between April 1st, 21:00 and April 2nd, 07:00
- Before April 1st, 21:00 and after April 2nd, 07:00 the differences are small, of the order of few meters.
Bad Broadcast Messages (resulting in satellite position errors up to about ±200km) were sent out starting April 1, 2014, 21h UT; Errors in geocentric, Earth-fixed x-coordinate of satellites of orbital plane 1. Normal transmission resumed at different times for different satellites. Complete “back to normal” around 7h UT of April 2.
Messages affected (from Hugentobler, E-mail April 4). Satellites 6, 7, 8, 14, 15, 23, 24 back to normal before end of April 1st (UTC).
The (R,S,W) coordinate system gives more insight into the structure of the 3-d position differences. For circular orbits the S component coincides with the along-track component. The position differences:
- are small in R (sub-km),
- are sizeable and almost constant in S within the three orbital planes,
- are sizeable and satellite-specific in the out-of-plane direction.
Bad Broadcast Messages (resulting in position errors up to about ±200km) were sent out starting April 1, 2014, 21:00; Errors in along-track coordinate S of satellites in orbital planes 1, 2, 3 are almost constant within each orbital plane. Normal transmission resumed at different times for different satellites. Back to normal around 7h UT of April 2.
Bad Broadcast Messages (resulting in position errors up to about ±200km) were sent out starting April 1, 2014, 21:00; Errors in out-of-plane coordinate W of satellites in orbital planes 1, 2, 3 are satellite-specific and periodic with amplitudes of the order of up to 200 meters. Normal transmission resumed at different times for different satellites. Back to normal around 7h UT of April 2.
Three orbits using satellite positions as pseudo-observations were determined for each satellite:
- F1_: GLONASS Positions from CODE final processing. Time interval April 1, 20h00m – April 2, 07h00m. Reference orbit (“truth”).
- GL_1: Positions from GLONASS Broadcast Ephemerides. Time interval April 1, 0h00m – April 1, 20h45m. (Correct GLONASS Broadcast Orbit).
- GL_2: Positions from GLONASS Broadcast Ephemerides. Time interval April 1, 21h00m – 23h45m. (Erroneous GLONASS Broadcast Orbit). Last observations of day skipped for satellites 6, 7, 8, 14, 15, 23, 24.
Each orbit was parameterized with 6 osculating elements (initial values) and 9 empirical parameters of the CODE model.
RMS values by satellite (plane 1):
| Satellite | RMS (F1_) | RMS(GL-1) | RMS(GL-2) |
|---|---|---|---|
| 1, plane 1 | 0.01 | 0.37 | 0.46 |
| 2, plane 1 | 0.02 | 0.34 | 0.42 |
| 3, plane 1 | 0.01 | 0.36 | 0.33 |
| 4, plane 1 | 0.02 | 0.40 | 0.32 |
| 5, plane 1 | 0.02 | 0.35 | 0.44 |
| 6, plane 1 | 0.01 | 0.33 | 0.23 |
| 7, plane 1 | 0.01 | 0.36 | 0.73 |
| 8, plane 1 | 0.02 | 0.37 | 0.44 |
RMS values by satellite (plane 2):
| Satellite | RMS (F1_) | RMS(GL-1) | RMS(GL-2) |
|---|---|---|---|
| 9, plane 2 | 0.03 | 0.31 | 0.30 |
| 10, plane 2 | 0.01 | 0.43 | 0.32 |
| 11, plane 2 | 0.03 | 0.31 | 0.47 |
| 12, plane 2 | 0.01 | 0.36 | 0.29 |
| 13, plane 2 | 0.03 | 0.36 | 0.35 |
| 14, plane 2 | 0.01 | 0.43 | 0.34 |
| 15, plane 2 | 0.02 | 0.58 | 0.40 |
| 16, plane 2 | 0.01 | 0.34 | 0.31 |
RMS values by satellite (plane 3):
| Satellite | RMS (F1_) | RMS(GL-1) | RMS(GL-2) |
|---|---|---|---|
| 17, plane 3 | 0.02 | 0.36 | 0.42 |
| 18, plane 3 | 0.01 | 0.60 | 0.36 |
| 19, plane 3 | 0.02 | 0.33 | 0.30 |
| 20, plane 3 | 0.02 | 0.39 | 0.29 |
| 21, plane 3 | 0.02 | 0.36 | 0.39 |
| 22, plane 3 | 0.01 | 0.34 | 0.80 |
| 23, plane 3 | 0.01 | 0.38 | 0.24 |
| 24, plane 3 | 0.01 | 0.35 | 0.40 |
The RMS values show:
- F1_: The consistency of the 15-parameter solution with the more elaborate CODE orbit model.
- GL-1 and GL-2: the positions of both GLONASS-derived positions obey the laws of Celestial Mechanics.
- The GL-1 and GL-2 positions cannot be mixed in orbit determination!
The GL-1 and GL-2 positions are compared to the F1_ positions by estimating three rotation angles about the equatorial inertial x-, y-, and z-axes.
| Solution | Rot x (“) | Rot y (“) | Rot z (“) | RMS (m) |
|---|---|---|---|---|
| GL-1 | 0.00 | -0.00 | 0.38 | 2.0 |
| GL-2 | -1763.25 | 5.28 | 2.09 | 70.7 |
Rotation of GLONASS Broadcast Orbits w.r.t. CODE 1-d orbits F1_. Erroneous GLONASS Broadcast positions show a consistent rotation of about 0.5 degrees about inertial x-axis.
Understood
- The GLONASS April 1st event was not a joke.
- The failure was caused by wrong Broadcast Ephemerides
- …probably caused by a software update, activated on April 1st at 9h p.m. UTC (24h/0h “Moscow Time”)
- The error was:
- identified by the GLONASS provider almost immediately after the activation of the new BE…
- but not communicated to the users of GLONASS.
- The correction (switch back to “old” software release?) took place satellite-by-satellite (starting on April 1, 10h45m p.m. UT).
- The effect on receivers obviously was dependent on the receiver type & firmware release.
Aspects to be considered by the IGS
- The Broadcast Ephemerides (BE) are not needed for the IGS post-processing activities (ultra-rapid, rapid, final).
- No actions needed concerning BE for these applications
- Report about malfunctioning receivers might be collected routinely and made available in a report.
- The BE are, however, of vital importance for the IGS Real-time Service (corrections w.r.t. BE are sent out); the BE are a “single point of failure” for the IGS real-time Experiment.
- Therefore one should either:
- implement an IGS real-time validation of all BE of all GNSS and make it available to the user community.
- or:
- reconsider the use of the BE for the IGS Real-time Experiment, as well.
- Therefore one should either:
Aspects to be considered by the IGS (cont)
GPS and GLONASS are today fully deployed systems. Despite that fact only combined solutions (or GPS-only solutions) are generated in the IGS.
One might apply the following principle for fully deployed systems in the IGS:
- Prior to combination, parameters of general interest (e.g., station coordinates, Earth Rotation Parameters) should be determined separately using only the observations of either GPS or GLONASS.
- GNSS-specific problems are more easily identified this way.
- The same principle should be observed by receiver software/firmware.
In addition, the quality of BE of all GNSS should be monitored by the IGS. For that purpose the predicted part of the IGS ultra-rapid orbits might be used.
Back-up Slides
Multi-GNSS Experiment (MGEX)
- MGEX call-for-participation released mid-2011 (ongoing)
- Steered by Multi-GNSS Working Group (MGWG)
- Some 27 contributing agencies from 16 countries
- Global tracking network, mostly real-time
- State-of-the-art receivers and antenna
- Tracking of Galileo, BeiDou, QZSS, SBAS (but no IRNSS, yet)
- Free and open access
- Data archives at CDDIS, IGN, BKG (RINEX 3.x)
- Real-time NTRIP caster (RTCM3-MSM)
- Product archive at CDDIS
IGS Real-time GNSS Service
Open Data:
- Observations & derived products freely available
- Streaming data over IP Networks in real-time
- Best effort operations, distributed governance
- Funded by national agencies, institutions, science
- Playing some global coordination role
Open Source:
- Supporting Real-time GNSS tools for Linux, Solaris, Window, Mac
- Multi-stream decoder, feeding GNSS engines, etc.
- Combining, encoding and uploading orbit/clock corrections
- Precise Point Positioning options
- Support of all GNSS through RINEX-3
Open Standards:
- Standardization in RTCM is understood as key issue
- Concepts and messages for all types of corrections
- Make PPP an optional alternative to Network RTK
The GLONASS April 1st Incident: a case study (AGNES network)
AGNES network of Swisstopo (one receiver type). Chart of “Processing performance” showing Release/Finish/Start times and # Stations over dates 24.03–05.04. Impact was heavy! GPS tracking was affected, as well!
Bad Broadcast Messages (resulting in position errors up to about ±200km) were sent out starting April 1, 2014, 21:00; figure shows errors in geocentric Earth-fixed y-coordinate of satellites of orbital plane 1. Normal transmission resumed at different times for different satellites. Normal transmission resumed around 6h UT of April 2.
Bad Broadcast Messages (resulting in position errors up to about ±200km) were sent out starting April 1, 2014, 21:00; figure shows errors in geocentric Earth-fixed z-coordinate of satellites of orbital plane 1. Normal transmission resumed at different times for different satellites. Normal transmission resumed around 6h UT of April 2.
Bad Broadcast Messages (resulting in position errors up to about ±200km) were sent out starting April 1, 2014, 21:00; Errors in radial coordinate R of satellites in orbital planes 1, 2, 3 are “only” of sub-km size. Normal transmission resumed at different times for different satellites. Back to normal around 7h UT of April 2.
Rotation of GLONASS GL-2 Broadcast Orbits w.r.t. CODE 1-d orbits F1_, separately for each orbital plane:
| Plane(s) | Rot x (“) | Rot y (“) | Rot z (“) | RMS (m) |
|---|---|---|---|---|
| 1 | -1762.83 | 5.40 | 2.06 | 56.1 |
| 2 | -1763.44 | 5.20 | 2.18 | 54.8 |
| 3 | -1763.47 | 5.81 | 2.04 | 81.4 |
| all | -1763.25 | 5.28 | 2.09 | 70.7 |
→ no plane-specific systematics.
The GLONASS Broadcast orbits GL-1 are consistent with the CODE precise orbits F1_ up to April 1st 2014, 20h59m UTC. All GLONASS Broadcast Orbits of type GL-2 show a rotation w.r.t. the CODE precise orbits F1_ within the time interval April 1, 2014, 21h00m – April 2, 07h00m UTC. The rotation takes place almost uniquely around the x-axis of the inertial equatorial coordinate system.
The GLONASS April 1/2 event was almost a “non-event” in the IGS, because:
- Only about 1/3 of the GLONASS tracking data were missing in the time slot April 1, 21h UT to April 2, 7h UT.
- only combined GPS/GLONASS products (and no GLONASS-only products) are generated,
- no highly time-resolved products are generated,
- the conventional IGS analysis is performed calendar day by calendar day (UT) and:
- only 3 hours of GLONASS data were affected on April 1st,
- only 7 hours of GLONASS data were affected on April 2nd.
The result of slide 17 was independently checked:
- The differences of the orbit normal vectors of the GLONASS orbital planes, using correct & corrupted messages, respectively, were decomposed into a rotation about the x-axis and one perpendicular to it.
- The following figure shows that only the rotation about the x-axis differs significantly from zero – and corresponds to the values obtained with the standard method.
(Chart: “Orbital plane rotation around x (red) and perpendicular (blue)” shows the x-axis rotation constant at about 0.49 degrees across all 24 satellite numbers, while the perpendicular rotation stays near zero.)
Epilogue: At the time of transition “back to normal” the satellites were visible (with an elevation of >20 deg) from at least one of the Command Tracking Sites on Russian Territory.