DELL ISTITUTO ITALIANO DI NAVIGAZIONE

Transcription

1 ATTI DELL ITITUTO ITALIANO DI IGAZIONE Rvsta d Navgazone Aerea, pazale, Marttma e Terrestre THI IUETHEME: Advanced, non commercal, GN Applcatons N 87 JANUARY-FEBRUARY 8 Poste Italane pa pedzone n Abbonamento Postale 7% DCB

2 Index G.Perrotta ro_: ann dal lanco 4 M.Ls putnk e la nascta della navgazone spazale 9 F.Tomasello La certcazone per l uso aeronautco della navgazone satelltare 6 G.Gragla, G.Guarno EGNO n the road doman: trals campagn results analyses 5 A.Pacco, A.Greco, M.Vultaggo RTK-EGNO:prove dnamche nel Golo d Napol n 5 navgazone marttma.falzn, M.Mnopol, A.Caporal, C.Morn, F.Pep, I. Adam,.Rozsa GALIlea: GALILEO Local Element Augmentaton or IE Predcton and ono/tropo Correcton 4 L.Marrad, G.Franzon, G.Pnell, A.Patrz The Galleo Ground Msson egment nonpr Reerence Recever Chan 54 B.Novello, M.Pn, F.Dovs, G.Marucco Intererence detecton n GN L band 65 A.Albanese, L.Marrad, L.Fogla, M.Ronztt The GRAIL UT oluton to GN Introducton nto ERMT/ETC 7 A.Ner, A.M.Vegn, A.D Nep TOA and DOA based localzaton servces n IEEE 8. networks 8 A.Angrsano, A.Pacco, M. Vultaggo Copertura della costellazone QZ (Quas Zenth atellte ystem) 9

3 GALILEA: GALIleo Local Element Augmentaton or IE Predcton and Iono/Tropo Correctons teano Falzn, Mararosara Mnopol, pace Engneerng.p.A Alessandro Caporal, Carlotta Morn, Tommaso Pep, CIA,Unversty o Padova Józse Ádám, zabolcs Rózsa, Department o Geodesy and urveyng, Budapest Unversty o Technology and Economcs, BUTE-HA Research Group or Physcal Geodesy and Geodynamcs ABTRACT The artcle descrbes new methods and algorthms to locally predct, montor and possbly mprove n near real tme the perormance o the postonng servce oered by the Galleo satelltes. The proposed applcaton can be consdered as a Galleo Local Element whch augments the global Galleo servce n a local area. In partcular ths paper descrbes prncples and algorthms to: Predct the gnal n pace Error (IE) n a local scenaro, mergng orbtal data wth real-tme observatons; Compute user poston dependent onospherc and tropospherc correctons. A local network o contnuously operatng reerence statons produces observaton data wth a hgh raw data update rate. These data have to be collected and processed together wth global normaton such as satellte ephemers data. The results such as ntegrty normaton, IE predctons, and onospherc as well as tropospherc correcton data have to be broadcast to local users whch are n proxmty o these statons, so to provde addtonal normaton or mprovng both the accuracy and the saety o the navgaton. Ths work has been perormed under the GALILEA (Galleo Local Elements Augmentaton) project, counded by GJU (Galleo Jont Undertakng). INTRODUCTION We address here how the GALIEA concept can be ully ntegrated n the Galleo scenaro. It s well known that each o Galleo satellte wll broadcast navgaton tmng sgnals together wth navgaton data sgnals. These sgnals wll contan not only the clock and ephemers correcton data essental or navgaton but also ntegrty sgnals whch provde a global space-based augmentaton servce. The ntegrty normaton can be dened as a measure o the trust that can be placed n the correctness o the normaton suppled by the system. Integrty ncludes the ablty o the navgaton system to provde users wth tmely and vald warnngs (alerts) when the system must not be used or the ntended operaton. For several applcatons related to road, avaton, ral, (tran control) and martme (harbour navgaton), the user requrements can be expressed n term o accuracy, tme to alarm, ntegrty rsk, contnuty, avalablty. In the ol (aety o Le) applcaton, the IE predcton n near real tme s an essental normaton, especally n crtcal condtons. Further, or the Open ervce applcaton n whch low cost sngle requency termnal are used, atmospherc correctons, n addton to derental correctons, could provde a sgncant added value to the standard servce. The proposed applcaton n the GALILEA project can be consdered as a Galleo Local Element whch augments the global Galleo servce n a local area. A network o local statons produces observaton data wth hgh rate. Data are collected, merged wth global normaton (e.g. satellte ephemers) and processed. Outputs are broadcast to local users va wreless Internet; they are composed by ntegrty normaton, as IE predcton, and onospherctropospherc correcton data. nce ths concept s restrcted at a local area, GALILEA can be consdered a Galleo Local Element whch can provde an augmentaton servce just or users who are n proxmty o reerence statons, lmtng, n ths way, the number o possble users. I the local concept can represent a lmtaton, on the other hand t enhances the qualty o the servce to be provded n a lmted regon, such as a cty, wth a hgh concentraton o users, nvolvng publc and prvate nterest, or cvl and mltary arports. GALILEA YTEM ARCHITECTURE The GALILEA system archtecture s composed by the elements whch are represented n Fgure. Reerence tatons. The statons are permanent and complyng wth the or EPN (European Permanent Network) standards as equpment and ste/monumentaton characterstcs. Each staton transmts typcally Hz raw code and phase data n real tme to the processng aclty. The local network approach envsoned n GALILEA has two major advantages: the transmtted onospherc and 4

4 tropospherc correctons can be user locaton dependent only a grd n space s dened; a IE predcton can be ndependent on sgnal errors relatve to a specc staton only a multple staton approach s selected. Meteo tatons. Local meteorologcal normaton (pressure, temperature, humdty) can sgncantly mprove the tropospherc modelng accuracy. Fgure : GN and Meteo tatons MA-L UMT/GPR (NTRIP) Internet EUREF GALILEA Rado Broadcast Faclty Rado Broadcastng (NTRIP) Iono/tropo IA valdty Tme /EUREF data Olne Integrty Montorng Perormance Assessment Processng Faclty Obs/Nav data Local Meteo Data ol Reerence taton Fgure : GALILEA Operatonal scenaro Meteo taton Processng Faclty. Ths aclty has to collect data rom derent sources and elaborate them. It s composed by: Error Predcton and Correcton Module: ths tool generates the IE predctons and the onospherc and tropospherc correctons that have to be transmtted to the user. Olne Integrty Montorng Module (IMM): the Olne Integrty Montorng computes the Integrty Rsk and/or the User Protecton Level n the local servce area based on the models and the local correctons that are used n the Processng Faclty. Rado Broadcast Faclty. Drect rado transmsson to ol users s accounted to a dedcated broadcast aclty, whle communcaton wth MA-L users s perormed usng the Internet. A network o 7 GN (belongng to the Hungaran Actve GP Network) statons was used or the valdaton campagn and a set o meteorologcal statons as close as possble to the GN statons was selected (Fgure ). The valdaton campagn was planned and perormed n collaboraton wth BUTE (Budapest Unversty o Technology and Economcs). The valdaton test was done both wth real (GP) observaton data and wth smulated (Galleo) data, usng the EA tool GF (Galleo ystem mulaton Faclty). mulated data allow the sotware to be tested n presence o a modeled satellte alure. THE ERROR PREDICTION AND CORRECTION MODULE The Error Predcton and Correcton Module (EPCM Tool) was developed by pace Engneerng and CIA, ts scope s to gve as output IE predctons or each satellte n vew and onospherc/tropospherc correcton coecents. The EPCM tool nputs can be dvded nto: taton dependent nput: o RINEX observaton les; o Meteorologcal data produced by a local Meteo taton; o INEX les whch provde precse staton coordnates ( avalable). atellte dependent nput: o orbts and clocks (ultra rapd predctons consdered totally relable, hal observed and hal predcted), and ERP (Earth Rotaton Parameters); o RINEX navgaton les. They contan clock ephemers models broadcasted by GN satelltes and used by the end user. Ther naccuraces relatve to the precse orbts and clocks are a prmary cause or the IE to exceed the IA. o IA les, part o the ntegrty message transmtted wthn the Galleo Open ervce. In our project IA wll be consdered as a constant (or each epoch and satellte) upper threshold whch should not be exceeded by the sgnal error (IE). The Fgure reports the Block Dagram o the EPCM Module. 4

5 INPUT taton dependent RINEX Observaton Fles INEX Fles Meteorologcal Data atellte dependent p Fles ( orbts, sat clock,..) RINEX Navgaton Fles IA Fles EPCM COMPUTATION MODULE Module EPM Tool Iono/ Tropo Delay Ionospherc/ Tropospherc Correcton IE Estmate Error Values Data Fuson Iono/Tropo correcton relatve to user poston IE Predcton Fgure : EPCM Block Dagram OUTPUT RTCM Moded Message IE Iono Predcton Correcton Tropo Correcton The EPCM Module s composed by: Module: ths module processes predcted orbtal les and RINEX navgaton les and t gves the OCE (Orbtal and Clock Error) as output; EPM Module: the Error Predcton Module computes the pseudorange resduals at each tme and or each satellte-staton combnaton, models them wth a RBTB (Range Bas Tme Bas) model and computes the nstantaneous error value at satellte level wth a multple staton approach. Ionospherc/Tropospherc Correcton Module: ths module computes the onospherc and troposherc delay at each tme epoch and or each satellte-staton combnaton and gves the correcton coecents as output, ater an nterpolaton. Ionospherc and tropospherc correctons are output n the orm o coecents: n ths way, the correctons are dependent on the user poston. Data Fuson Module: ths module computes the IE predcton by mergng the Module and EPM Module outputs. The ollowng sectons descrbe the algorthms and models mplemented n the EPCM. THE MODULE The purpose o the module s to estmate the gnal n pace Error due to Orbt and Clock Errors (OCE) as an ndex o the gnal In pace Error (IE) based on data, through the comparson between the ephemers contaned n the sp les and the ephemers computed rom the orbtal parameters ncluded n each navgaton message. We compute, at each nstant, the satellte coordnates usng the and RINEX navgaton les, and project ther derence ( x, y, z) onto the lne connectng the user locaton and the satellte poston comng rom the determnaton. Fnally, to obtan the OCE, the satellte clock error s added to the projecton. Denotng wth ( x, y, z ) the user coordnates; ( x, y, z ) the satellte coordnates computed rom the data and ( x y, z ), the ones calculated rom the RINEX navgaton les, we show n Fgure 4 the vectors, whle ormulas are reported n Eq. and. r r r r (Eq. ) ( x x ) + ( y y ) + ( z z ) ( x x ) + ( y y ) + ( z z ) ( x, y, z ) r X X r r r ( x, y, z ) r r Fgure 4: Orbtal and Clock Error X ( x, y, z ) r r r r r r cosθ (Eq. ) r The OCE value s gven by: ( r) + ( t T ) c OCE (Eq. ) There s a problem o samplng tme: whle the ephemerdes are avalable every 5 mnutes, the IE computed wth the RBTB model presents a samplng tme o ew second. In order to compare the two values t s necessary to nterpolate the data to compute a Hz IE: ater a trade o analyss a cubc splne nterpolator was chosen. The Fgure 5 reports the OCE plot, as dened n Eq., relatve to a wndow o only one day. We have vered that the peak n the gure was caused by a lack o vald ephemers n the RINEX navgaton les. Ths caused an old ephemers be propagated beyond ts valdty tme untl a new one was avalable. The use o old ephemers resulted n an OCE whch was promptly detected by our sotware. 4

6 OCE [m OCE, V epoch [s] OCE [m OCE ZOOM o the dscontnuty area, V Fgure 5: OCE trend or V epoch [s] THE IONO/TROPO CORRECTION MODULE The Iono/Tropo Correcton Module ncludes a onospherc and a tropospherc correcton module. The onospherc module s composed by two submodules. The rst submodule computes, or each reerence staton and satellte, the ono-ree pseudoranges and the coecents to correct the rst and second (only three requences are avalable) order reracton eects, both n the two (GP) or three (Galleo) requences case. uch computaton s done or every set o observables contaned nto the RINEX observaton les o each staton, the number o whch s strctly dependent on the samplng tme o the RINEX les themselves. The outputs o the rst submodule act as nputs or both the EPM module and or the second submodule, whch computes the coecents o the onospherc grd nterpolatng the data rom all the reerence statons. uch coecents allow the user to compute the onospherc eect or each pont o the workng area once that the observaton ste longtude and lattude are known. Lke the onospherc module, the tropospherc module can be dvded nto two submodules. The rst computes, or each staton o the network and or each satellte, the tropospherc path delay combnng the data rom the meteo staton (pressure, temperature and relatve humdty) once the satellte elevaton s known. The satellte elevaton s only dependent on the staton coordnates and the satellte coordnates. The tropospherc path delays act as nputs or both the EPM module and the second tropospherc submodule, whch, analogously to what happens or the second onospherc submodule, computes the coecents o tropospherc grd, nterpolatng the data rom all the reerence statons. uch coecents allow the user to 5 55 OCE compute the tropospherc path delays or each pont o the workng area once that the observaton ste longtude and lattude are known. The Fgure 6 reports the Iono/Tropo Module block scheme. In the ollowng subparagraphs the two model descrptons are reported. The Ionospherc Module Ater a trade o prelmnary analyss we have concluded that the best choce to acheve a better accuracy n the results, n order to correct the reracton eects, s to ollow an analytc approach, usng the double/trple requency technques to correct the rst/second order reractve eects, rather than a mathematcal model. INEX OB METEO Iono module Iono delay Tropo module Tropo delay Loop over satellte Loop over satellte Loop over staton Loop over staton Loop over epoch Loop over epoch COMPUTE IONO GRID COMPUTE TROPO GRID a ono, b ono, c ono, d ono a tropo, b tropo, c tropo, d tropo Error Predcton Module To the user To the user Fgure 6: Iono/tropo module block scheme n EPCM In the onosphere the reracton ndex depends strctly on the electron densty and s proportonal to the nverse o the square o the carrer requency. In ts most smpled orm t can be wrtten as : a a a n p (Eq. 4) where a, a, a are coecents. When GP sgnals penetrate through the onosphere, the range error o the transmsson paths due to the varaton o the reractve ndex can be expressed as : ( n ) δρ ds (Eq. 5) s ubsttutng Eq. 4 nto Eq. 5 yelds: a a a δρ ds + ds + ds (Eq. 6) 4 s s s For dual requency satelltes lke GP or GLONA only the rst two terms n Eq.6 need to be consdered: 44

7 a a A A δρ + ds + (Eq. 7) s In the case o the GP, where only two requences st ( L and L ) are present, only the order onospherc reractons can be elmnated applyng a dual-requency technque, as reported n Eq. 8: a A δρ ds (Eq. 8) s An ono-ree pseudorange ρ ono ree s computed: ρ ono ree where ρ L γρ L γ ρ L and (Eq. 9) ρ L are the pseudorange measured on L and L channel respectvely and L γ, wth L 7.6 MHz. L MHz and L Wth Galleo two derent approaches are possble. The rst s a least square model based on the search o the optmal values o the unknown parameters ( A and the onoree pseudorange) whle the second s an analytcal approach analogous the one developed or the two requences, and whch computes also second order reracton eects. We have decded to choose the last approach because t merges optmally the smplcty o the algorthm wth a hgh degree o precson. o t possble to consder the Eq. 7 where (,,) represents the generc Galleo requency. The rst order onospherc reracton removed combnatons can be obtaned rom Eq. : dρ dρ dρ dρ A A A A A A From these equatons t results Eq. : (Eq. ) ρ ( ) ρ ( ) A ( ) + ( ) + ( ) ρ ( ) ρ ( ) A ( ) + ( ) + ( ) (Eq. ) where ρ j ρ - ρ j. ettng equal to zero, ths equaton becomes exactly the expresson well known or the two requences case. Once that the value o ρ ono-ree, (,,) s known, t s possble to compute the pseudorange correcton due to onospherc eects or each pont o the workng area, knowng the lattude and the longtude o the observaton ste. In order to have an accurate model, we need to cover the entre workng area wth a sucent number o statons. The scenaro used to test our grd s exactly the same scenaro as the valdaton campagn: the workng area s characterzed by sx Hungaran reerence statons, coordnates o whch are ncluded nto the INEX le gven to the onosphercal module as nput. The basc dea s to nterpolate, wth a surace, the values o the onospherc correcton obtaned or the reerence statons o the local area. From the computaton o the surace parameters (coecents o the polynomal nterpolaton), t s possble to calculate the onospherc correcton (ρ ono-ree, ) relatve to each pont o the plane. Frst o all we transorm the cartesan coordnates ncluded nto INEX les nto geodetc ones. Then ater computng the barycentre o the statons network, we project the geodetc coordnates nto a plane. In the plane coordnate system, wth orgn n the network barycentre (wth label ), the coordnates o the -th ste are, to rst order: n ( ϕ ϕ ) RE (Eq. ) e ( λ λ ) RE cos( ϕ ) where ϕ and λ are the coordnates o the barycentre, ϕ and λ are the coordnates o the -th staton, and R E s the Earth radus ( R E 67km); alternatvely, a more exact cartographc representaton such as the UTM cylndrcal projecton can be used. Now we are ready to wrte the nterpolaton polynomal. Ater a prelmnary analyss n whch we have taken nto consderaton the resduals and the magntude order o the coecents n relaton to ther estmated uncertanty, we have decded to stop the polynomal to the thrd degree. takng nto account hgher order terms dd not sgncantly mprove the accuracy. An example o the mathematcal model used to compute the 45

8 pseudorange correctons rom the polynomal coecents s: d ρono (Eq. ) ( ϕ, λ) a + b ϕ + c λ + d ϕ + e λ +.. ono ono ono where ϕ and λ represent, respectvely, the longtude and the lattude o the user locaton n a specc tme t. The a ono, b ono, c ono, d ono and e ono are the polynomal coecents. They result rom the nterpolaton o the onospherc delays relatve to the reerence statons, the coordnates o whch are well known. In Fgure 7 we have plotted the onospherc delays relatve to the workng area. The gure shows the contour lnes n D and the same lnes projected onto the n-e plane, and the postons o the reerence statons. It s possble to observe that the hgher values o the onospherc delays are acheved n correspondence o the boundary zones, regons where the accuracy o the coecents obvously degrades. Fgure 7: Ionospherc grd. Plot o the onospherc delays Fgure 8 represents the general approach ollowed n mplementng the onospherc (tropospherc) delays nterpolaton. A uture step o our work could be the merger o the data relatve to derent satelltes computed at zenth n order to provde a general ndex o the onospherc (tropospherc) actvty n the local area. The Tropospherc Module Ater a trade o analyss we have concluded that the model whch best suts the requrements addressed n our work, n terms o precson and accuracy, s the Langley and Collns model [Re. 9]. It provdes a derent parametersaton or the wet and dry part o the atmosphere, gong over what was the lmt o the Hopeld model, and, at the same tme, t doesn t lack n precson or reduced elevaton angles, as happens n the aastamonen model. ono ono Fgure 8: cheme o the approach To compute the tropospherc delays or all the elevaton angles and not only delays at the zenth, t s necessary to ntroduce a mappng uncton. Our am was to nd a model or the mappng uncton whch provdes good perormance beng sutable or the majorty o the elevaton angles, and, at the same tme, whch s as smplest as possble accordng wth real tme mplementatons n a computaton lmted recever. For these reasons, and also n conormty wth the model adopted by EA, we have decded to use the Black and Esner mappng uncton [Re. 8]. The Langley and Collns model s based on the classcal Daves expresson [Re. 9] reported n Eq. 4: t tropo ( d dry where m( ) + dwet ) m( El ) c (Eq. 4) El s a mappng uncton dependng on the satellte elevaton and c the speed o lght. The Langley and Collns model rewrtes the tropospherc delays as reported n Eq.5: d d dry wet g R d β βh zdry T ( λ+ ) g βh Rd β zwet T where g s the acceleraton o gravty, (Eq. 5) Rd s the gas constant or the dry ar, h s the geodetc heght o the recever n meters, T s the temperature n K at the sea level, β s the temperature lapse rate n K/m, λ z are s the lapse rate o the water vapour, zdry and wet the delays due respectvely to the dry and wet part o 46

9 the atmosphere computed n correspondence o the π zenth E. At the sea level ( h ) the expressons or the delays become as reported n Eq. 6: z z dry wet 6 m krd P g m g k R 6 e ( λ + ) βr T d d (Eq. 6) where k and k are the reractvty constants. The expresson o the mappng uncton s reported n Eq. 7:. m( El ) (Eq. 7).+ sn ( El ) vald or El > 5. As or the onospherc pseudorange correcton, t s possble to grd the tropospherc delays o the reerence statons. For each observaton epoch we calculate polynomal coecents whch, nserted nto a mathematcal model (Eq. 4), provdes the tropospherc pseudorange correcton or each locaton nsde the dened workng area. The predcted troposherc pseudorange correcton s: d ρ tropo (Eq. 8) ( ϕ, λ) a + b ϕ + c λ + d ϕ + e λ +.. tropo tropo tropo In Fgure 9 we report the trend o the tropospherc delays patterns computed at the same epoch consderng the same satellte, smlarly to Fg.7 or the onosphere. tropo tropo Fgure 9: Tropospherc delays pattern THE EPM MODULE The Error Predcton Module computes the pseudorange resduals at each tme and or each satellte-staton combnaton, ts them wth a moded RBTB model and computes the nstantaneous error value at satellte level wth a multple staton approach. Pseudorange resduals are obtaned applyng the error correctons to the pseudorange measurements and subtractng the satellte-staton dstance based on the navgaton message. Ionospherc and tropospherc errors are acqured rom the Iono/tropo correcton module, satellte clock bas s calculated usng the parameters contaned n the navgaton message, staton clock bas s computed perormng a sngle pont soluton at each tme step. Even ater correctons have been appled, pseudorange resduals are stll ar rom showng a regular trend descrbng the IE. A resdual error due to onosphere, troposphere and clock remans ater the correctons; recever nose and some components o the multpath error whch typcally behave as a random uncton cannot be corrected. In order to cancel the random eects on the pseudorange resduals and to denty the correct trend o the error, a least square t on a speced data nterval s necessary. The moded RBTB model has resulted rom the model tradeo analyss to be the best opton or the pseudorange resduals least square t. It allows to cancel the random eects on the pseudorange resduals and to t rapd error varatons thanks to the thrd equaton term tme dependence. The moded RBTB model equaton has the ollowng orm o Eq. 9: DRB ρ RB + + TB() t (Eq. 9) sn ( El) where RB s a constant range bas term manly due to the radal component o the ephemers error and partly due to the other (tangental, cross track) components o the ephemers error and to a constant satellte clock bas. The derental range bas DRB ncludes the error terms not o orbtal orgn correlatng wth the elevaton, as tropospherc resdual errors and part o the multpath error. The tme bas TB term ncludes error components whch correlate wth the tme: ths s the derence respect to the RBTB model, where the thrd term s dependent rom the satellte radal velocty. Ths choce s due to some consderatons about the IE behavor: typcally, ephemers and clock errors tend to lnearly grow wth tme snce the last navgaton message update tme. Ths s due to the act that orbtal parameters and 47

10 clock correcton data contaned n the navgaton message represent a sort o predcton whch becomes more and more naccurate as tme ncreases. The model has been tested usng data smulated by the tool GRANADA on a satellte-staton bass, selectng a t nterval o 6 seconds. mulated data nclude onospherc and clock errors wth related correctons, multpath and recever errors; a resdual tropospherc-lke error has been ntroduced to model uncertantes n the correcton algorthms. Pseudorange resduals are obtaned as derence between the corrected measured pseudorange and the known dstance between staton and satellte, whch s aected by the errors contaned n the navgaton message. ample tme has been set to second. Frstly, let us consder the model behavour when no satellte alure s ntroduced n the smulaton, as reported n Fgure. The last value o the t curve represents an estmaton o the error (composed by IE and resdual errors) at the actual tme. In order to smulate the operatve condtons o the sotware, the t data wndow s moved n a certan tme rame usng a step tme o 6 seconds. It s possble to reconstruct the nstantaneous error trend n a certan tme nterval (see Fgure ). The eect o the rapd oscllatons n the error trend wll be attenuated as the error s averaged wthn a number o reerence statons, snce ths eect does not depend on the satellte error components. Now, let us ntroduce an addtonal error startng rom a certan nstant smulatng a satellte alure, and consder the nstantaneous error trend), as reported n Fgure. IE [m] pseudorange resduals RBTB t t [sec] Fgure : Moded RBTB model t IE [m] IE [m] nstantaneous error trend t [sec] Fgure : Instantaneous error trend nstantaneous error trend real IE t [sec] Fgure : Instantaneous error trend n case o satellte alure THE DATA FUION MODULE The data uson module acqures the IE estmaton rom the module and the nstantaneous error value rom the EPM and t computes the IE correcton to be appled to the IE estmaton usng the nstantaneous error trend. Fnally, t computes a IE predcton to be output to the user. The Fgure represents a block scheme o the Data Fuson Module. It s mportant to underlne that snce a IE correcton has to be appled to the IE estmaton, t does not have to take n account IE varatons already descrbed by the estmaton. Hgh requency varatons o the nstantaneous error trend caused by the random multpath dstrbuton have to be reduced, whle low requency varatons o the nstantaneous error trend caused by the elevaton angle dependency have to be gnored. 48

11 Module IE estmaton 5 nstantaneous error ltered error 4 Error Predcton Module Instantaneous error Reconstruct the error trend Data Fuson Module [m] Compute IE correctons Compute IE predcton Loop over satellte IE predcton Fgure : Data Fuson Module block scheme Frst o all the IE predcton has to be subtracted to the Instantaneous Error Trend; n ths way, the remanng error can be consdered as a correcton to the IE. A properly desgned lter, whch generates IE correctons derent by zero only n case the IE s really varyng respect to the module predcton, was selected. A recursve (or Innte Impulse Response) dgtal lter has been selected n order to perorm the hgh and low pass operaton on the sgnal. Recursve lters use all the prevous sgnal samples to generate the ltered sgnal at the present tme. They show a very good behavour n the tme doman and are easy to mplement. Hgh perormances n the requency doman are not necessary or ths applcaton: thereore, a sngle pole recursve lter has been selected. The Fgure 4 shows the lter output, consderng an error trend generated by a sngle staton, whle Fgure 5 shows the same error trend when a satellte alure here modeled as a cubc uncton occurs. The ltered error trend shows the ollowng eatures, whch make t sutable to consttute a IE correcton term: Its value s around zero when no unpredcted IE varatons occur. It rapdly ncreases when unpredcted IE varatons occur. In order to compute the IE correcton value at the present tme and to predct ts trend, a second order polynomal t has to be appled to a speced tme nterval o the ltered error. When no unexpected IE varaton occurs, the correcton term wll be close to zero, thus the IE predcton wll be close to the predcton. A long t tme nterval wll ncrease the predcton accuracy, but wll also make t slower to predct ast IE varatons due to a satellte alure. [m] t [sec] Fgure 4: Fltered error trend nstantaneous error ltered error t [sec] Fgure 5: Fltered error trend n presence o a satellte alure VALIDATION CAMPAIGN TET REULT Valdaton wth smulated data In order to valdate the proposed algorthms pseudorange observatons were needed on three requences. To acheve ths, the Galleo ystem mulaton Faclty V. (VEGA IT Gmbh, Germany) has been used. Although GF s capable to smulate Galleo and GP sgnals as well, only Galleo satelltes were used or our smulatons. The raw data generaton was done usng the Galleo Nomnal Constellaton o 7 Galleo satelltes, and a vrtual Galleo recever were placed exactly at the poston o the BUTE permanent GP staton n Hungary. The vrtual Galleo recever tracked the E5a, E5b and Lb data contnuously. All the smulatons were done n the tme rame o 6-- : to 6-- :, wth the observaton nterval o one second. 49

12 To valdate the OCE module o the EPCM, three derent scenaros were modelled. The rst scenaro was a smple reerence model, n whch no addtonal error (alure) had been ntroduced. The second data set contaned a slowly ncreasng clock error or PRN 6, where the addtonal clock error ncreased rom µs to. µs n mnutes (correspondng to a clock drt o 5.55E- s/s), whle the thrd data set contaned a slowly ncreasng clock error where the same lnear change occured n mnute (correspondng to a clock drt o 5.55E- s/s). In the rst smulaton, the alert o a IA valdty tme lower than s occurs ater s rom the satellte clock alure start tme. Ths tme nterval can be consdered as a vald measure o the IE predcton reacton tme; nevertheless, the ltered error reacts much aster to the error occurrence, but not enough to produce a warnng n the output. Fgure 6 represents the IE predcton ater a ew seconds rom the alure start. One can observe that the IE predcton s too slow to track a rapdly ncreasng error, reachng the value o m ater one mnute. uch an error s drectly detected by EPCM ater a ew seconds rom the alure start on the bass o the pseudorange resduals: thus, a relable IE predcton s uneasble and unnecessary. In the second smulaton, the satellte clock drt s ten tmes slower respect to the prevous case. The reacton tme, as prevously dened, s 67 s. Fgure 7 represents the IE predcton trend or the second smulaton. One can observe that, n ths case, the real error s tracked (and overestmated) by the IE predcton. 4 Fltered error IE predcton Real error IE [m] Fltered error IE predcton Real error t [s] Fgure 7: IE predcton, second smulaton Valdaton wth real data In order to valdate the concept wth real data, GP and meteorologcal data has been collected usng the ground network shown on Fgure. Valdaton o IE predcton usng data For the rst EPCM run, a complete set o observaton/navgaton real data rom 7 GP statons and predcted orbtal/clock data wth a 6 hours latency tme have been used. The observaton wndow s rom 7//7 :: to 7//7 :: (one hour), and the observaton sample tme s one second. For the second EPCM run, the same observaton/navgaton set has been used, whle the le has been replaced by measured orbtal/clock data. Fgure 8 represents the EPCM output IE estmaton compared wth the Orbtal and Clock Error obtaned wth both predcted and measured data or V 7. IE [m] pred OCE IE real OCE -6-8 IE [m] t [s] Fgure 6: IE predcton, rst smulaton t [s] Fgure 8: Error trend, V 7 5

13 For each satellte, the predcted OCE standard devaton respect to the real OCE and the IE standard devaton respect to the real OCE (IMA) have been computed on the entre observaton perod. The results are summarzed n Table : V OCE std IMA average Table : OCE standard devaton and IMA values Wth the excepton o the V9, or whch the predcted OCE s qute derent respect to the real one, the IMA value s largely below the expected maxmum value (. m); urthermore, n three cases (V9, V6, V9), the IMA value s smaller than the OCE standard devaton value. In all cases, the IMA value s close to the OCE standard devaton. The man outcomes o ths test case are that, n absence o a satellte alure:. The OCE standard devaton s, n 5 over 6 cases, largely below the expected maxmum value.. The IMA value results to be very close to the OCE value. Pont has the mportant mplcaton that predcted orbtal/clock data can be ecently used as reerence to compute the IE. Pont mples that the EPCM output IE s a relable measure o the Orbtal and Clock error at satellte level n absence o a satellte alure. Ionospherc and tropospherc correctons grd valdaton Ths test ams at measurng the accuracy o the ono/tropo correctons nterpolaton grd n the operatonal area. An addtonal two requences GP recever has been placed n a pont o the operatonal area, not concdent wth a reerence staton and close to an exstng meteo staton. For the rst EPCM run, a complete set o 7 reerence statons has been used (see Fgure ). The second EPCM run has been perormed usng only one staton, the addtonal recever labeled OMZ. Ths recever s placed near the center o the operatonal area. The observaton wndow s rom //7 :: to //7 :: (one hour), and the observaton sample tme s one second. Table and Table show the EPCM tropo and ono correctons computed n the OMZ locaton, the real values computed usng the OMZ data and the derence between them, whch s a measure o the correcton accuracy. As expected, the delta values or the tropospherc correctons are n the order o a ew centmeters, whle the delta values or the onospherc correcton are n the order o 5 meters. V EPCM correcton OMZ correcton delta Table : EPCM tropo correcton accuracy V EPCM correcton OMZ correcton delta Table : EPCM ono correcton accuracy Ths test clearly gves evdence that the nterpolaton o the onospherc correctons rom a set o reerence statons cannot produce nterestng results provded that the recevers have derent behavours n measurng observatons on two requences. On the contrary, the EPCM tropospherc correcton accuracy s, or all the satelltes n vew, largely below the expected value. CONCLUION The proposed local archtecture or the GALILEA project has resulted to be capable to provde an added value to the baselne Galleo servce, n terms o ntegrty and accuracy. Galleo standard ntegrty parameters, as IA and Integrty Flag, can be complemented by addtonal normaton, as IE predcton, or users operatng n a local area. Ths normaton can be broadcast usng the wreless nternet by means o the NTRIP system. Three 5

14 ntegrty montorng concepts were analyzed olne to assess the capablty o local archtectures to mprove the ntegrty montorng perormance and to alert the users n case o a hazard. The man actvtes o the GALILEA project have been the development o error correcton modules and ther ntegraton and valdaton, the assessment o the ntegrty montorng perormance and the study o a data dssemnaton system. The project conclusons are summarzed n the ollowng. The EPCM sotware has been developed wth the objectve to provde to local users ntegrty and accuracy normaton, mergng global predcted orbtal/clock data wth local observaton and meteo data rom a set o reerence statons located n a 6 km radus operatonal area. Durng the models development phase, a tradeo analyss has been perormed or the onospherc and the tropospherc correcton models, takng n account the avalablty o local meteorologcal measures and the Galleo three requences. The RBTB t model has been moded ntroducng a tme dependent thrd term, whch makes t more ecent to t rapd error varatons. A new model has been developed to compute the Orbtal and Clock Error (OCE) rom the comparson between predcted satellte poston/clock data and navgaton message ephemeredes. The desgn and mplementaton o a dgtal recursve (IIR) lter has solved the data uson problem, where range measurements aected by a number o error sources have to be used to correct the accurate, but slow, -based error predcton. From the valdaton test results, the IE predcton algorthm mplemented n EPCM has proved hs capablty to provde the expected perormances wth both real and smulated nput data. Wth smulated data, t has been possble to test the IE predcton n presence o a satellte alure. Addtonally, the orbtal/clock predcted data have resulted to be well suted to be merged wth local observaton data n order to predct the IE. The use o local meteo data to compute the tropospherc delay has receved a postve eedback rom the valdaton tests. The correcton coecents, resulted by the tropospherc delay D nterpolaton, have shown to mprove the user poston accuracy n the operatonal area. The onospherc delays computed by derent reerence statons have resulted to be qute uncorrelated, makng the nterpolaton process superluous. Due to the raw data generaton tool characterstcs, the three requences onospherc correcton technque could not be ecently valdated wth the provded smulated data. Consderng the tght schedule mposed or models development and sotware ntegraton, the EPCM perormances are more than satsactory. The merge o global ( orbtal and clock predctons) and local (observaton) data n a sngle ntegrty normaton (IE predcton) has proved to be a easble concept. The output IE predcton, compared wth the Galleo broadcast IA, s able to provde to the user an normaton about the IA valdty tme. The nterpolaton o the ono/tropo correctons n the local area has shown the capablty to mprove the user poston accuracy provded that a certan degree o spatal correlaton s shown by the correctons: durng the valdaton tests, ths condton held or the tropospherc correctons. However, the sotware capabltes and perormances can be mproved wth urther work ocused at models, operaton and valdaton. Concernng models, the use o phase measurements has been proposed by CIA n order to mprove the onospherc model accuracy. Concernng operaton, the development o a real-tme verson o the sotware would be advsable, n order to progress towards an operatve tool and allow the test o the real-tme dssemnaton capablty. Concernng valdaton, an extended valdaton campagn s hghly recommended: t may nclude the use o pseudoltes and the custom to a specc applcaton envronment, as aeronautc or tran. ACKNOWLEDGMENT The GALILEA project s co-undend by the Galleo Jont Undertakng under the contract number: GJU/5/4/CTR/GALILEA The GALILEA consortum s composed by pace Engneerng.p.A. (Italy), Center o tudes and Actvtes or pace G. Colombo (Italy), NavPos ystem GmbH (Germany), Budapest Unversty o Technology and Economcs (Hungary) and Federal Agency or Cartography and Geodesy (Germany). ACRONYM BUTE Budapest Unversty o Technology and Economcs CIA Centro Interdpartmentale tud e Attvtà pazal DRB Derental Range Bas EPN European Permanent Network EA European pace Agency EPCM Error Predcton Correcton Module EPM Error Predcton Module ERP Earth Rotaton Parameters GALILEA GALIleo Local Element Augmentaton GN Global Navgaton atellte ystem GP Global Postonng ystem 5

15 GF Galleo ystem mulaton Faclty Internatonal GP ervce IIR Innte Impulse Response MA-L Medum Accuracy Low aety NTRIP - Networked Transport o RTCM va Internet Protocol OCE Orbtal and Clock Error RBTB Range Bas Tme Bas RINEX Recever INdependent EXchange ormat INEX oluton Independent Exchange Format IA gnal In pace Accuracy IE gnal In pace Error IMA gnal-in-pace Montored Accuracy ol aety o Le td tandard Devaton V atellte Vehcle TB Tme Bas UTM Unversal Transverse Mercator REFERENCE [] Global Postonng ystem: Theory and Applcatons, Vol. I, B.W. Parknson, J.J. plker, AIAA [] The Galleo Integrty Concept and Perormance, V. Oheler, F. Luongo, H. L. Trautenberg, J. Boyero, J. Krueger, T. Rang, Galleo Industres GmbH [] GN Local Component Integrty Concepts, C.. Dxon, Journal o Global Postonng ystems,. [4] Global Postonng ystem, Theory and Practce, B.Homann- Wellenho, H. Lchtenegger, J. Colln, prnger Wen New York. [5] Classcal Electrodynamcs, J.D. Jackson, John Wley and sons edtor. [6] Handbook o Geophyscs and the pace Envronment, centc Edtor A.. Jursa, Ar Force Geophyscs Laboratory, Ar Force ystems Command, Unted tates Ar Force, 985. [7] Trple-Frequency Method or Hgh-Order Reractve Error Modellng n GP Modernzaton, Z. Wang, Y. Wu, K. Zhang, Y. Meng, Journal o Global Postonng ystem, Vol. 4, No -:9-95, 5. [8] A New Tropospherc Propagaton Delay Mappng Functon or Elevaton Angles Down to, Jmng Guo, chool o Geodesy and Geomatcs, Wuhan Unversty, Chna, Rchard B. Langley, Department o Geodesy and Geomatcs Engneerng, Unversty o New Brunswck, Canada. [9] Lmtng Factors n Tropospherc Propagaton Delay Error Modellng or GP Arborne Navgaton, Paul Collns and Rchard Langley, Geodetc Research Laboratory, Department o Geodesy and Geomatcs Engneerng, Unversty o New Brunswck, Canada. James LaMance, Y Corporaton, Colorado prngs, CO, U..A. teano Falzn teano Falzn s a techncal manager at pace Engneerng. In 99 he receved a dploma n Aeronautcal Engneerng. Hs background ranges rom Earth Observaton msson studes and products to the satellte navgaton studes and algorthms denton, passng through msson analyss and AOC /W. Recently, n the GN eld, he has been nvolved n Galleo ystem Test Bed (GTB) development actvtes and GJU research contracts. Mnopol Mararosara Mararosara Mnopol receved a Master Degree n pace Engneerng rom the Unversty o Naples Federco II n Aprl 6. he s nvolved n the GALILEA project or GJU and n the development o uture applcatons or GALILEO. Alessandro Caporal Alessandro Caporal holds a degree n Physcs (magna cum laude) at the Unversty o Padova (975) and a Ph.D. n Physcs rom the Ludwg Maxmlan Unversty n Munch, Germany, and the Max Planck Insttute uer Physk und Astrophysk (979). He has done postdoctoral research at NAA - Marshall FC, Alabama, at the Department o Earth and Planetary cences o MIT, Massachusetts, and NAA - Goddard FC, Maryland. He s actvely nvolved n the European Permanent Network o GP statons and s currently proessor o old Earth Physcs at the Unversty o Padova. zabolcs Rózsa zabolcs Rózsa holds an Mc n urveyng and Geomatcs Engneerng (998), and an Mc n Busness Admnstraton (4). He earned hs PhD n Earth cences at the Budapest Unversty o Technology and Economcs n. He worked at the unversty o Karlsruhe as research asocate between and 4. He ocused on precse GP postonng and ts applcaton to the deormaton analyss o tectonc eatures o the Rhne-Graben. nce 4 he s an assstant proessor at the Department o Geodesy and urveyng at the Budapest Unversty o Technology and Economcs. Hs research area s precse GP postonng and ts applcatons or deormaton analyss and meteorology. 5