T. Sakai, K. Matsunaga, K. Hoshinoo, ENRI T. Walter, Stanford University T. Sakai, K. Matsunaga, K. Hoshinoo, ENRI T. Walter, Stanford University Prototype of SBAS and Evaluation of the Ionospheric Correction Algorithms Prototype of SBAS and Evaluation of the Ionospheric Correction Algorithms ION NTM 2006 Monterey, CA Jan , 2006

ION NTM Jan Sakai, ENRI S LIDE 1 Implementation of the prototype of SBAS:Implementation of the prototype of SBAS: –A prototype of SBAS has been successfully implemented; –Outputs the complete SBAS messages; tested with the SBAS user receiver simulator; –The overall performance is comparable with the MSAS. Evaluation of ionospheric correction algorithms:Evaluation of ionospheric correction algorithms: –Using the above prototype as an evaluation tool; –Evaluation of the current algorithm: ‘Planar Fit’; the storm detector trips a lot during storm ionospheric condition; –Proposed algorithm with the zeroth order fit reduces the protection levels to the third part of the current algorithm. Introduction

ION NTM Jan Sakai, ENRI S LIDE 2 MSAS is now under operational test procedures for IOC:MSAS is now under operational test procedures for IOC: –Protection levels are hugely conservative with large margins; not reflecting the actual performance; –Needs reducing protection levels to improve availability; However, MSAS has no ‘Testbed’:However, MSAS has no ‘Testbed’: –There is only the operational system; cannot be used for testing new algorithms; –The prototype of SBAS will be a powerful tool for evaluation of new algorithms for improvement of MSAS. QZSS is being developed in Japan; also needs a testbed for investigation of wide-area augmentation technique.QZSS is being developed in Japan; also needs a testbed for investigation of wide-area augmentation technique. Motivation

ION NTM Jan Sakai, ENRI S LIDE 3 Actually computer software running on PC and UNIX:Actually computer software running on PC and UNIX: –‘RTWAD’ written in C language (not MATLAB, sorry); –Consists of the essential components and algorithms of WADGPS; –Developed for study purpose; does not meet the safety requirements for civil aviation navigation facilities. Generates the complete SBAS messages:Generates the complete SBAS messages: –Outputs one message per second; –250-bit message without FEC encoding; –Optional output in NovAtel $FRMA format; works as direct input to SBAS user receiver simulator. Implemented Prototype

ION NTM Jan Sakai, ENRI S LIDE 4 $FRMA,272, ,138,80811EA4,250,53081FFDFFDFFDFFDFFDFFDFFDFFDFFDFFD FFDFFDFFFBBBBBBBBBBBBAC1CD280*7C $FRMA,272, ,138,80811EA4,250,9A0C1FFDFFDFFDFFDFFDFFDFFDFFDFFDFFD FFDFFDFFFBBBBBBBBBBBBB7E76F80*0F $FRMA,272, ,138,80811EA4,250,C661FFDFFDFFDFFDFFDFFFBBBBB CD8A40*70 $FRMA,272, ,138,80811EA4,250,5306FFBFFFF B963FC0*0D $FRMA,272, ,138,80811EA4,250,9A091FFDFFDFFDFFDFFDFFDFFDFFDFFDFFD FFDFFDFFFBBBBBBBBBBBB806D3340*77 $FRMA,272, ,138,80811EA4,250,C60D1FFDFFDFFDFFDFFDFFDFFDFFDFFDFFD FFDFFDFFFBBBBBBBBBBBB924AAE40*08 $FRMA,272, ,138,80811EA4,250,5361FFDFFDFFDFFDFFDFFFBBBBB FE640*73 $FRMA,272, ,138,80811EA4,250,9A61FFDFFDFFDFFDFFDFFFBBBBB8A D00*05 $FRMA,272, ,138,80811EA4,250,C60A1FFDFFDFFDFFDFFDFFDFFDFFDFFDFFD FFDFFDFFFBBBBBBBBBBBBA6BE8CC0*03 $FRMA,272, ,138,80811EA4,250,530E1FFDFFDFFDFFDFFDFFDFFDFFDFFDFFD FFDFFDFFFBBBBBBBBBBBBA99E5040*0A Message Type ID (6 MSBs) Preamble Message Length SBAS Satellite PRN Time CRC Message Sample

ION NTM Jan Sakai, ENRI S LIDE 5 Currently running in offline mode:Currently running in offline mode: –Used for various evaluation activities; –RINEX observation files are input as monitor station observations; provided from GEONET continuous observation network operated by GSI, Japan; –Thus the distribution of monitor stations is variable and the historical ionospheric storm events can be tested. Utilizes only code phase pseudoranges on dual frequencies:Utilizes only code phase pseudoranges on dual frequencies: –Needs no carrier measurements; –Outputs one message per second although RINEX is 30- second sampling. Implemented Prototype

ION NTM Jan Sakai, ENRI S LIDE 6 User Receiver Simulator SBAS user receiver simulator:SBAS user receiver simulator: –Also software running on PC and UNIX; –Processes RINEX observation file with L1 pseudorange; carrier smoothing applied; –Decodes SBAS message stream (in NovAtel $FRMA records) and apply them to the observations; –Tested with WAAS and MSAS messages. The implemented prototype was evaluated by this simulator:The implemented prototype was evaluated by this simulator: –With GEONET observations at some locations; –Output messages work functional; –Evaluates position accuracies and protection levels.

ION NTM Jan Sakai, ENRI S LIDE 7 Monitor Stations We used GEONET sites as monitor stations.We used GEONET sites as monitor stations. Dual frequency observation sampled every 30 seconds.Dual frequency observation sampled every 30 seconds. 6 monitor stations similar to the MSAS.6 monitor stations similar to the MSAS. 5 user locations for evaluation.5 user locations for evaluation. GEONET for Monitor Stations GEONET for User Stations MSAS Service Area

ION NTM Jan Sakai, ENRI S LIDE 8 Example of User Position Error Example of user positioning error at Site (Takayama; near center of monitor station network).Example of user positioning error at Site (Takayama; near center of monitor station network). Period: July 2004; active ionosphere condition.Period: July 2004; active ionosphere condition. Standalone GPS Augmented by the Prototype HorizontalErrorVerticalError m m m m System StandaloneGPS m m m m Prototype RMS Max RMS Max

ION NTM Jan Sakai, ENRI S LIDE /11/14-16 HorVer 2004/7/22-24 HorVer 2004/6/22-24 HorVer 2005/11/14-16MSAS HorVer RMSMax RMSMax RMSMax RMSMax RMSMax Site North South Performance (Nominal) Unit: [m]

ION NTM Jan Sakai, ENRI S LIDE 10 Performance (Severe Storm) 2004/11/8-10 HorVer 2003/10/29-31 HorVer RMSMax RMSMax RMSMax RMSMax RMSMax Site North South Large errors due to ionosphere.Large errors due to ionosphere. Users still protected within protection levels.Users still protected within protection levels. Unit: [m]

ION NTM Jan Sakai, ENRI S LIDE 11 Protection Levels (Quiet) Protection level Protection level of MSAS Actual error Protection levels during quiet ionosphere at site (second southern user).Protection levels during quiet ionosphere at site (second southern user). Protects users with large margins.Protects users with large margins. MSAS provided further conservative protection levels.MSAS provided further conservative protection levels.

ION NTM Jan Sakai, ENRI S LIDE 12 Protection Levels (Storm) Protection level Actual error Protection levels during storm ionosphere at site (second southern user).Protection levels during storm ionosphere at site (second southern user). Still protects users with large margins.Still protects users with large margins. However protection levels grow large; this means low system availability.However protection levels grow large; this means low system availability.

ION NTM Jan Sakai, ENRI S LIDE /11/14-16 HorVer 2004/7/22-24 HorVer 2004/6/22-24 HorVer 2005/11/14-16MSAS HorVer Site NominalIonosphere Protection Level Statistics 2004/11/8-10 HorVer 2003/10/29-31 HorVer Site RMS of protection levels in meters.RMS of protection levels in meters. Grows large for storm ionospheric conditions.Grows large for storm ionospheric conditions. Severe Storm Ionosphere

ION NTM Jan Sakai, ENRI S LIDE 14Problems WADGPS Corrections work well:WADGPS Corrections work well: –Positioning accuracy is m horizontal and m vertical, respectively, over mainland of Japan; –The largest vertical error was less than 40 meters; could support APV-I operation (HAL=40m, VAL=50m). Protection levels are hugely conservative:Protection levels are hugely conservative: –HPL and VPL completely protected users; –However, protection levels were not reflecting the actual performance regardless of ionospheric activities; –Needs reducing protection levels to improve availability. Investigates this problem using the prototype system.Investigates this problem using the prototype system.

ION NTM Jan Sakai, ENRI S LIDE 15 VPL during Storm Vertical protection levels during storm ionosphere at site (second southern user) with the baseline algorithm.Vertical protection levels during storm ionosphere at site (second southern user) with the baseline algorithm. Most of VPL came from ionospheric component.Most of VPL came from ionospheric component. To reduce protection levels, the primary issue is ionosphere.To reduce protection levels, the primary issue is ionosphere. Protection level Ionospheric component Actual error

ION NTM Jan Sakai, ENRI S LIDE 16 UIVE and Actual Residual UIVE (user ionospheric vertical error) is interpolated from GIVE (grid ionospheric vertical error).UIVE (user ionospheric vertical error) is interpolated from GIVE (grid ionospheric vertical error) UIVE works as ionospheric component of protection levels.5.33 UIVE works as ionospheric component of protection levels. Large margin even during historical severe storm.Large margin even during historical severe storm UIVE Actual ionospheric residual

ION NTM Jan Sakai, ENRI S LIDE 17 Without the storm detector algorithm, UIVE was computed like this.Without the storm detector algorithm, UIVE was computed like this. The large UIVE in daytime is resulted in by trip of storm detector.The large UIVE in daytime is resulted in by trip of storm detector. The actual ionospheric residual exceeded 5.33 UIVE only once even without storm detector.The actual ionospheric residual exceeded 5.33 UIVE only once even without storm detector UIVE Actual ionospheric residual UIVE without Storm Detector

ION NTM Jan Sakai, ENRI S LIDE 18 The ionospheric storm detector caused a lot of false alert conditions lowering system availability:The ionospheric storm detector caused a lot of false alert conditions lowering system availability: –When storm detector trips, the associate GIVE value is set to the maximum. To avoid such a problem there are two possible ways:To avoid such a problem there are two possible ways: –Develop an alternative safety mechanism instead of the storm detector; –Develop a method to compute GIVE values instead of setting to the maximum when storm detector trips. Here we introduce the latter algorithm.Here we introduce the latter algorithm. Storm Detector Problem

ION NTM Jan Sakai, ENRI S LIDE 19 0-th order fit (1 parameter) Ionospheric delay Rmax 1-st order fit (3 parameters) Estimated delay at IGP We can reduce the order of fit when the storm detector trips; the planar model cannot be applied.We can reduce the order of fit when the storm detector trips; the planar model cannot be applied. Only one parameter needs to be estimated; equivalent to the weighted average.Only one parameter needs to be estimated; equivalent to the weighted average. Let us see UIVE and the actual residuals induced by the zeroth order fit.Let us see UIVE and the actual residuals induced by the zeroth order fit. Introducing Zeroth Order Fit

ION NTM Jan Sakai, ENRI S LIDE 20 UIVE by Zeroth Order Fit UIVE computed with the zeroth order fit without the storm detector algorithm.UIVE computed with the zeroth order fit without the storm detector algorithm. UIVE is larger than planar fit.UIVE is larger than planar fit. The largest residual was within 5.33 UIVE even during the historical storm events; the zeroth order fit does not need the storm detector.The largest residual was within 5.33 UIVE even during the historical storm events; the zeroth order fit does not need the storm detector UIVE Actual ionospheric residual

ION NTM Jan Sakai, ENRI S LIDE 21 The zeroth order fit works well and protects residuals within 5.33 UIVE even during storm ionospheric conditions.The zeroth order fit works well and protects residuals within 5.33 UIVE even during storm ionospheric conditions. Thus the following adaptive algorithm can be employed:Thus the following adaptive algorithm can be employed: –1. Apply the standard planar fit with the storm detector; –2. If storm detector does not trip, employ resulted correction and GIVE; –3. Otherwise, or the number of IPPs is insufficient for the standard planar fit, perform the zeroth order fit. This algorithm will reduce the number of IGPs with the maximum GIVE due to trip of storm detector.This algorithm will reduce the number of IGPs with the maximum GIVE due to trip of storm detector. Adaptive Algorithm

ION NTM Jan Sakai, ENRI S LIDE 22 Protection Levels (Storm) Reduced protection levels to the third part; improved availability.Reduced protection levels to the third part; improved availability. Still protects users with large margins.Still protects users with large margins. Baseline Algorithm Adaptive Algorithm

ION NTM Jan Sakai, ENRI S LIDE 23 GIVE Statistics Baseline algorithm Adaptive algorithm Current baseline algorithm produced the maximum GIVE (GIVEI=14) for 50% of IGPs.Current baseline algorithm produced the maximum GIVE (GIVEI=14) for 50% of IGPs. The adaptive algorithm reduced the maximum GIVE conditions and replaced to GIVEI=13.The adaptive algorithm reduced the maximum GIVE conditions and replaced to GIVEI=13. GIVEI = 15 Not Monitored GIVEI = 14 Maximum GIVE

ION NTM Jan Sakai, ENRI S LIDE /11/8-10 HorVer 2003/10/29-31 HorVer Site BaselineAdaptive BaselineAdaptive BaselineAdaptive BaselineAdaptive BaselineAdaptive Reduction of Protection Levels RMS protection levels in meters during storm ionospheric conditions.RMS protection levels in meters during storm ionospheric conditions. The adaptive algorithm reduced protection levels to the level of third part of the baseline algorithm.The adaptive algorithm reduced protection levels to the level of third part of the baseline algorithm.

ION NTM Jan Sakai, ENRI S LIDE /11/ /10/ % 81.7 % 29.2 % 62.8 % 37.4 % 88.2 % 26.1 % 65.9 % 39.4 % 83.3 % 15.8 % 36.8 % % 69.6 % 20.0 % 34.3 % % 34.6 % 14.0 % 18.8 % Site BaselineAdaptive BaselineAdaptive BaselineAdaptive BaselineAdaptive BaselineAdaptive LNAV/VNAV Availability 2004/11/ /10/ % 77.6 % 28.4 % 59.0 % 36.9 % 86.8 % 25.9 % 62.1 % 39.1 % 81.6 % 15.8 % 34.9 % 38.3 % 68.7 % 19.9 % 33.9 % 25.7 % 33.8 % 14.0 % 18.7 % APV-I (LPV) Availability System Availability

ION NTM Jan Sakai, ENRI S LIDE 26 Realtime operation:Realtime operation: –For implementation and tests integrity functions (TTA); –RTWAD runs with causality to input observations; little modification for realtime operation; –Signal biases will be computed day by day; –ENRI is installing realtime monitor stations; 6 stations up to now and one more shortly; Evaluation activities in offline mode:Evaluation activities in offline mode: –Effects of additional monitor stations (IGS stations); –Testbed for dual frequency SBAS. Upcoming Plans

ION NTM Jan Sakai, ENRI S LIDE 27 Realtime Stations We have already installed 6 stations with realtime datalink to ENRI, Tokyo.We have already installed 6 stations with realtime datalink to ENRI, Tokyo. Additional station in Toyama is to be installed shortly.Additional station in Toyama is to be installed shortly. MSAS Stations Realtime Statins MSAS Service Area

ION NTM Jan Sakai, ENRI S LIDE 28 Experimental Equipment Trimble 4000SSE in Sapporo NovAtel MiLLennium and IP converters at ENRI, Tokyo

ION NTM Jan Sakai, ENRI S LIDE 29 A prototype of SBAS has been successfully implemented:A prototype of SBAS has been successfully implemented: –Overall accuracy: m horizontal and m vertical; –Protection levels completely protects users. Evaluation of ionospheric correction algorithms using prototype:Evaluation of ionospheric correction algorithms using prototype: –The current algorithm caused a lot of the maximum GIVE; –Adaptive algorithm will reduce the protection levels to the third part of the current algorithm and improve availability. Future works will include:Future works will include: –Realtime operation of prototype system; –Simulation of dual frequency SBAS. Contact: Conclusion

ION NTM Jan Sakai, ENRI S LIDE 30 Ionospheric Delay: Quiet

ION NTM Jan Sakai, ENRI S LIDE 31 Ionospheric Delay: Storm

ION NTM Jan Sakai, ENRI S LIDE 32 MSAS Architecture