ZE SZ Y T Y N A U K O W E II IN T E R N A T IO N A L C O N FE R E N C E ________________ PO LITEC H N IK I ŚLĄ SK IEJ 2002 T R A N SPO R T SY ST E M S T E L E M A T IC S T S T '02 T R A N SP O R T z.45, nr kol. 1570
GPS, DGPS, GSM, GPRS, TCP/IP, DGPSIP Libor S E ID L 1
U SIN G O F G SM -G PR S N E T W O R K A N D IN TE R N E T FO R D IF F E R E N T IA L G PS (GNSS)
The expansion o f Internet connectivity by GSM-GPRS network to transport elements allows new possibilities of differential GPS (GNSS) techniques and applications. This paper presents several ideas of effective DGPS protocol design intended for UDP packet layer of GSM-GPRS communication channel.
U Ż Y T K O W A N IE SIE C I G SM -G PR S O R AZ IN TE R N E TU W R Ó Ż N IC O W Y M SY ST E M IE GPS
Rozwój komunikacji GSM-GPRS pozwala na powstanie nowych możliwości zastosowań technik GPS. Zakłada się, że łączenie z Internetem poprzez GSM-GPRS lub inne sieci radiowe będzie standardem dla elementów wyposażenia transportu. Wydatki na taki system zależą głównie od całkowitych pojemności kanału danych - przekazywanych lub otrzymywanych. Aktualnie na świecie nie jest określony standard zalecanego pasma oraz protokołu danych dla kanału GPRS. Dlatego w referacie przygotowano odpowiedni protokół przesytu danych. Osiągalny w praktyce współczynnik kompresji jest 10-15 większy w porównaniu do binarnego formatu danych, a 5-50 jedno-sekundowe obserwacje mogą być zapisane w jeden pakiet.
1. IN TR O D U C TIO N
T he G PS is w ell know n as a global navigation satellite system . T he analogous system s are the R ussian G LO N A SS and G A LILEO system prepared by EU. A ctually, the GPS is the only one fully applicable global satellite position determ ination system on the w orld.
T he reachable GPS perform ance for m ost o f civil users (Standard position service - SPS) is lim ited; therefore, it is not sufficient for som e applications. In these years, the perform ance lim its are caused by native m easurem ent errors (ephem eris error, satellite clock error, ionospheric and tropospheric refraction, m ultipath effect and receiver im perfection).
The system precision for SPS users can be significantly reduced by technique know n as Selective A vailability (SA). T he SA is suspended at present, b u t it is liable to political and military decision. B esides, the system integrity is insufficient for m any o f life safety applications.
T he considerable im provem ents in accuracy and integrity are accessible by differential m easurem ent m ethods (D G PS, D GNSS). T hese im provem ents arise because the largest m easurem ent errors are strongly correlated over distance and vary slow ly w ith tim e. The
1 Czech Technical University, Faculty of Electrical Engineering, Dept, of Radioelectronics Technickd 2. 166 27 Praha 6, Czech Republic, seidl@fel.cvut.cz
66 L ibor SEIDL differential system accuracy is lim ited by residual errors only, i.e. m ainly by residual atm ospheric signal refraction and m ultipath effect.
T he differential techniques are based on processing o f m easurem ent records from twice o f GPS receivers - user receiver and reference receiver. C onsequently, a data com m unication channel is a necessary part o f each real-tim e differential application. A t present the General Packed R adio Service (G PRS) o f G SM netw ork appears as a perspective link with relative sufficient technical and econom ical param eters. T he non-existence o f appropriate DGPS data standard for the G SM -G PR S channel appears as other barrier for im m ediate usage. The requirem ent analysis and draft o f suitable protocol is a topic o f this work.
2. A NALYSIS O F R EQ U IR EM ENTS
Since 2 nd M ay 2000, the SA was turned off. Due to this, position accuracy for GPS SPS users increases to 8-25 m (95% , horizontal). In im plication o f this, the DGPS users structure changed. T he autonom ous GPS accuracy becam e sufficient for many o f applications with m ean accuracy requirem ents (e. g. navigation on road in urban areas). A t present, DGPS techniques are used in applications w ith high accuracy o r integrity requirem ents.
DGPS protocols w ithout carrier phase inform ation (e.g. R T C M SC 104, m essage 1 or 9) are suitable for standard position accuracy only (1-5 m). The absence o f carrier phase inform ation causes a lim itation o f m ultipath effect suppresses. F or accurate applications, the carrier phase inform ation is necessary. B esides, it is desirable to prepare extensions for new expected signal com ponents (GPS L2-C /A , L5).
T he zero level o f SA causes a change o f tim e characteristics o f GPS m easurem ent errors. T he tim e correlation and dynam ics o f these errors depend considerable on actual state o f ionosphere. T he m ean period o f DGPS data retransm ission m ay be extended significantly, but it has to be adaptive according to im m ediate requirem ents (depend on m om entary state o f ionosphere, etc.). T he DGPS correction may be extrapolated on the receiver side in continuance o f data retransm ission interval. T he extrapolation error w ill be m onitored in the reference station, because this station know s correction values, w hat w ere sent to this user. As soon as extrapolation error overflow s given lim it, the new correction set may be sent im m ediately.
O perating expenses o f G SM -G PR S com m unication depend on total received or transm itted data capacity. T herefore, the DGPS data protocol needs ensure the m inim al cum ulative data volum e that is the m inim al average data bandw idth.
A special m essage or other protocol instrum ent has to be prepared fo r w ell-tim ed integrity w arnings.
3. A N A LY SIS O F C U R R EN T DGPS D A TA STA N D A R D S
T he D G PS system s can be classified according to form o f data transm itted through a com m unication channel betw een the reference station and the user equipm ent. T he first group o f DGPS system s uses corrections, w hich are com puted in reference station as a difference betw een m easured and expected pseudoranges. C arrier phase m easurem ent is rare included.
T he o ther system s transm it observations by com m unication channel - code pseudoranges and carrier p hase m easurem ents. T he advantage o f this m anner is sim plicity, accuracy and
Using o f G SM -G PR S netw ork and internet for differential GPS (GNSS) 67 independence on satellite ephem eris. T he main disadvantage is a greater volum e o f transm itted data. T his disadvantage can be suppressed by effective data form atting.
D ata protocol recom m ended by R TC M SC104 [2] is the m ost spread standard for DGPS in real tim e applications. T he m ost o f code-DG PS receivers is com patible w ith RTCM m essages o f types 1 o r 9. Special m essages (18-21) w ere designed for m ore precise techniques (R TK ), b u t these are used rare. T hese m essages contain carrier phase m easurem ents, both pseudorange and correction forms are supported. Insufficient parity protection will be replaced by 24-bit CR C protection in future [4].
As a m ain standard for postprocessing (i.e. n o t real-tim e) D G PS purposes is used RINEX (R eceiver Independent Export) created by the A stronom ical Institute o f U niversity in Bern. It is a readable text form at verified during a long period o f tim e. T here is a certain possibility for extension and adaptation o f this form at for the future. R IN EX is not generally suitable to transm ission in real-tim e radio link due to a huge data volum e. C R IN EX (C om pact RINEX, packed form o f RIN EX) created by Yuki H atanaka in 1996 reduces the data volum e three tim es approxim ately. R IN EX and CR IN EX are very spread data form ats for postprocessing, b u t they are not suitable for real-tim e applications. T he fundam ental idea o f CRINEX com pression is interesting for real-tim e data com pression. W e use this idea in our protocol design.
A s a good exam ple o f data protocol intended for high accuracy real-tim e m easurem ent (RTK) may be declare C M R or CM R+ (C om pact M easurem ent R ecord Form at) designed by Trim ble. A lthough it is a proprietary protocol, it is open for public. T his protocol is optim al for the m inim al latency at appropriate channel with lim ited throughput, but it is non-optim al for G SM -G PR S channel, w hich requires m inim al cum ulative data volum e.
A t last years, the Internet data channel starts to be used for DGPS purposes. In 2000 W olfgang R upprecht im plem ented the first version o f DG PS Internet server D G PSIP. This technique is based on data transm ission in R T C M -SC 104 form at through T C P or UDP connection. D ata protocol o f D G PSIP program s available in this tim e (Jul 2002) is relative sim ple and not effective enough. M ore sophisticated protocol version is under developm ent.
DGPSIP is usable for G SM -G PR S connection too, but data bandw idth efficiency is not very good for actual protocol version.
4. ID EA AN D REA LISA TIO N O F DGPS PR O T O C O L D ESIG N
T w o protocol variants o f com pressed data o f GPS observations w ere defined. T he first one is m ore general and adaptable w ith good com press ratio, but its com press algorithm is com plicated and represents considerable num erical load. D esign o f the protocols is not com pleted yet.
T he second protocol is prim ary designed for G PS receiver G arm in G PS-35 (but it is independent on particular receivers type) and contains phase and C /A code observations on GPS LI band. T his protocol is optim ised for static o r slow ly m oving user o r reference station.
For its m ain advantage - im plem entation sim plicity - the algorithm was designed for sm all eight-bit m icrocontrolers. T he protocol is intended for raw pseudoranges data transm ission from m obile client to server in reference station. T he core o f redundancy reduction algorithm expresses F ig .l.
68 L ibor SEID L
T he input o f this algorithm is a separate satellite observation set in discrete tim e k, w hich contains:
• C /A code m easurem ent (pseudorange) C, (k) recalculated to carrier w avelength,
• C arrier integrated phase m easurem ent L t ( k ) ,
• Signal to noise ratio SN R t (k ) as a signal quality indicator.
As a first step the “code-phase” difference X, (k ) = C, (k) - L, (k ) is com puted. By this step a correlation betw een C, (A:) and L t (k) m easurem ents is reduced.
D ata volum e can be reduced significantly by quantization. T he acceptable inform ation loss w as assigned by observation record analysis and w as verified by quantization im pact on accuracy and reliability o f resultant position inform ation
w here function floor() sym bolises truncating to low er integer value and the quantization intervals arc specified for Garm in GPS-35 (A is L I w avelength):
As the next step, the tim e series decorrelation can be applied. T his technique is very w ell know n from C R IN E X com pression. It is based on com putation o f tim e series difference.
L (k ) = floorfL, (k ) / A L ] ■ A L X (k) = floor[X, (/c)/AXj- AX
S N R (k) = floor[SM?, ( k ) /A S N R ] - A S N R
( 1)
A L = — — , AX = — , A SN R = 3 [d B ].
1024 4 (2)
d L (k ) = L (k ) - L (k -1 ),
d 2L (k ) = dL (k ) - dL(k - 1 ) = L (k ) - 2 L (k - 1 ) + L (k - 2),
d }L (k ) = d 2L{k) - d L { k - 1) = L{k) - 3L (k - 1 ) + 3L (k - 2) - L (k - 3).
(3)
T he tim e series difference for X, (k) and SN R t (k) are com puted analogically.
Using o f G SM -G PR S netw ork and internet for differential GPS (GNSS) 69
T he final operation is m essage form atting. S eparate satellite observation coding is provided at first. F ive m essage form ats w ere designed; each form at has specific range o f values and order o f tim e differences. T h e FO form at contents full inform ation w ithout relation on previous m essages. T his form at is intended for initialisation o r singular events. T he F 3 or F4 form ats (bandw idth saving form ats) are usable in com pression process steady state. T he second or third tim e series difference d 2X ( k ) , cPL{k) are used in these form ats for redundancy saving.
70 L ibor SEIDL
11100000 Cycle slip S N R (k) L (k ) X { k )
[ ID - 8 bits 1 bit 5 bits 42 bits 40 bits
1101 Cycle slip d S N R (k) d L (k ) d X ( k )
1 ID - 4 bits 1 bit 3 bits 32 bits 16 bits
1100 Cycle slip dSN R (k) d 2L (k ) d 2X ( k ) ID - 4 bits l bit 3 bits 16 bits 8 bits
10 Cycle slip d 2L (k ) d 2X ( k ) (condition: d S N R (k) = 0 ) ID - 2 bits 1 bit 13 bits 8 bits
0 d 3L (k ) d 2X ( k ) (conditions: d S N R (k) = 0 , none cycle slip) ID - 1 bit 10 bits 5 bits
Fig.2. Message formats for effective observations saving
T he set o f single satellite m essages from one tim e is follow ed by tim e mark. The satellite list is inserted in event o f list change only and it is represented by bit mask.
O bservations from several seconds (5-50) are com pleted into one U D P packed and sent from client to reference station server. T he first observation in U D P packed uses form at F 0 for acquirem ent o f independence on previous packed.
T h e protocol design represents a com prom ise betw een efficiency and robustness. The robustness is needed by virtue o f UD P channel irresponsibility. T he possibility o f retransm ission o f dropped data is ensured by challenges sent through the Internet/G PR S backw ard channel. Therefore, the error-correcting coding is not necessary; the UD P protocol error detecting is adequate. O ne set o f observations (one second) for tw elve satellites represents approxim ately 250 bits in steady state. T his fact signifies a considerable data volum e saving com pare to original binary data from GPS receiver (see T a b .l).
Table 1 Compress efficiency of designed algorithm (Results of verification tests, receiver Garmin GPS-35, size o f IP and PPP heads and checksums is not included)
M essage form at E xperim ent 1 (static) Experim ent 2 (kinem atics)
Relative count F0 (%] 1.17 3.06
Relative count FI [%] 1.95 2.44
Relative count F2 [%] 5.93 4.51
Relative count F3 f%l 19.84 49.60
Relative count F4 [%1 71.1 40.39
O riginal record size [byte] 6 280 000 670 675
C om pressed record size [byte] 494 719 50 782
C om press factor 12.69 13.2
U sing o f G SM -G PR S netw ork and internet for differential GPS (GN SS) 71
5. C O N C LU SIO N
T he G SM -G PR S channel can be used for DGPS data transm ission. T he suitable data protocol w as designed in this work. In contrast to the R TC M protocol (m essages 1 o r 9) this protocol contains raw code and carrier observations. O ne set o f observations (12 satellites) is described by only about 250 b it (32 bytes) at in steady state. 5-50 observations (seconds) may be stored into one U D P packed. T he average com press factor for typical binary observation protocol (Garm in G PS-35) is b etter that 10. T his results in significant reduction o f operation expenses in DG PS applications w ith G SM -G PR S data channel in real-tim e (or nearly-real- time, because o f U D P packed com pleting delay).
D esigned protocol is intended for „inverse D G P S ” applications. T he raw m easurem ents (pseudoranges) from a m obile client are delivered to a server in DGPS (D G N SS) reference station. T he position inform ation is com puted in the reference station (nearly in real tim e) and may be sent back to m obile client. P osition inform ation m ay be used and logged in reference station server and the backw ard channel may be used fo r com m ands and processed navigation inform ation (e. g. in graphic form - as m ap or actual railw ay plan).
W e plane to continue in the protocol specification and optim isation in the future. T he application o f this w ork w ill be im plem ented on the D G PS/D G N SS reference station on the D epartm ent o f R adioelectronics at Czech T echnical U niversity in Prague, at Faculty o f Electrical Engineering.
BIB LIO G R A PH Y
[1] EKBERG P., RTK Compression forD A R C . [Thesis]. Linköping, 1999.
[2] RTCM Paper 11-98/SC 104-STD, RTCM Recommended Standards for Differential GNSS (Global Navigation Satellite Systems) Service, Version 2.2. Washington D.C., Radio Technical Commission for Maritime Services 1998.
[3] YAN Y „ Differential GPS Corrections Protocol for the Caltrans TCFI Automatic Vehicle Location System Testbed. http://www.cs.ucsb.edu/-almeroth/cla.sses/193/S01-van/
[4] ZEBHAUSER B.E., EULER J„ KEENAN C.R., A Novel Approach for the Use of Information from Reference Station Networks Conforming to RTCM V2.3 and Future V3.0, Leica Geosystems AG, Switzerland 2002.
R eviewer: Prof. Edw ard H rynkiew icz
