Seamless train localization based on BDS/INS/odometer/MM multi-sensor navigation system

Transcription

1 Associatio Iteratioal Global Navigatio Satellite Systems IGNSS Coferece 2016 Colombo Theatres, Kesigto Campus, UNSW Australia 6 8 December 2016 Seamless trai localizatio based o BDS/INS/odometer/MM multi-sesor avigatio system Si-Rui Che School of Electroics ad Iformatio Egieerig, Beijig Jiaotog Uiversity, P.R. of Chia @bjtu.edu.c Wei Jiag School of Electroics ad Iformatio Egieerig, Beijig Jiaotog Uiversity, Beijig Egieerig Research Ceter of EMC ad GNSS Techology for Rail Trasportatio, P.R. of Chia weijiag@bjtu.edu.c Bai-Ge Cai School of Electroics ad Iformatio Egieerig, State Key Laboratory of Rail Traffic Cotrol ad Safety, BJTU. Beijig Egieerig Research Ceter of EMC ad GNSS Techology for Rail Trasportatio, P.R. of Chia bgcai@bjtu.edu.c Jia Wag School of Electroics ad Iformatio Egieerig, Beijig Jiaotog Uiversity, Beijig Egieerig Research Ceter of EMC ad GNSS Techology for Rail Trasportatio, P.R. of Chia wagj@bjtu.edu.c Wei Shaggua School of Electroics ad Iformatio Egieerig, Beijig Jiaotog Uiversity, Beijig Egieerig Research Ceter of EMC ad GNSS Techology for Rail Trasportatio, P.R. of Chia wshg@bjtu.edu.c ABSTRACT BeiDou Navigatio Satellite System (BDS) has bee widely applied i areas icludig commuicatio ad trasportatio, emergecy rescue, public security etc., makig remarkable beefit i social ad ecoomic aspects. The itegratio of BDS ad Iertial Navigatio System (INS) has meat that it gives greater access to a more cosistet ad accurate positioig capability tha ever before. However, durig the log BDS outages, the performace of the itegratio system degrades greatly because of the characteristics of the Iertial Measuremet Uit (IMU) the sesor error would be accumulated with time whe it operates i a stadaloe mode. I this paper, a BDS/INS/odometer/Map-matchig (MM) positioig methodology for trai avigatio applicatios is proposed ad evaluated by the real test coducted i the wester regios of Chia. This method is proposed to solve the problem of trai positioig durig BDS outages whe the trais pass through the sigal obstructed areas, like bridges, tuels ad valleys. The seamless

2 trasitio of the trai operatio i various scearios ca be therefore maitaied. Whe the trai operated i the ope-sky eviromets, the BDS sigals are available to provide accurate positioig, the BDS/INS are itegrated to correct the INS errors by BDS measuremets ad calculate the velocity by odometer i avigatio frame. Meawhile, the itegratio system are also delivered accurate positioig measuremets i high updated rate. I the case of BDS sigals blocked, the itegrated system ca be seamless switched to the INS/odometer itegratio mode, the itegrated system corrects the INS errors by usig odometer measuremets so as to provide cotiuous ad acceptable positioig performace. I additio, MM techology is applied to improve the precise. I order to evaluate the proposed system, a real experimet was coducted i the wester regios of Chia. The experimetal results idicate that the proposed system ca provide the accurate ad cotiuous positioig service i both ope-sky ad BDS obstructed eviromets. KEYWORDS: Beidou Navigatio Satellite System; INS; odometer; mapmatchig; seamless trai localizatio 1. INTRODUCTION The scale of railway etwork has bee further expaded i Chia i recet years. With the developmet of the implemetatio of the Chiese Wester Developmet Strategy, which help acceleratig the social ad ecoomic developmet of the whole wester regio, the ceter of Chiese railway costructio has bee movig to the wester regio. The railway etwork costructio i wester regio is also uder quick developmet. Because of the vast territory ad sparse populatio, the railways i Chiese wester regio have the characteristics of log mileage, small traffic volume ad iadequate ifrastructure. The majority of these lies are desiged as sigle track railway, ad defied as Low Desity Lies (LDL). Oe represetative LDL is Chiese Qighai-Tibet railway, which locates i Chiese Tibet provice ad have more tha 300 km depopulated zoe at 4000 m above sea level. It is quite ecessary to implemet the trai positioig methodology with high reliability ad less maiteace for the low desity lies like Qighai-Tibet railway. The scee of Qighai-Tibet railway is show i left plot of Fig.1, ad the right plot shows the example of tuel i Qighai- Tibet railway. Figure 1. Qighai-Tibet Railway The Global Navigatio Satellite System (GNSS) ca improve the performace of the trai positioig system, ad has attracted icreasig attetio i recet years. It is able to reduce the

3 cost of the trai positioig system greatly by decreasig the trackside equipmets ad the maiteace workload, ad fulfil the special requiremets for safety ad efficiecy. As BeiDou Navigatio Satellite System (BDS) has bee able to fully cover the Asia-Pacific area, it has bee applied i may fields icludig commuicatio ad trasportatio, emergecy rescue, public security etc. i Chia, ad makig remarkable beefit i social ad ecoomic aspects (Soderholm et al, 2016). BDS has bee applied to achieve accurate avigatio ad precise positioig as well. However, similar to Global Positioig System (GPS) system, BDS receivers must track more tha four satellites i order to yield the precise coordiates. The more abudat resources of satellites, the better performace of positioig. Therefore the itegratio of BDS ad INS is able to give greater access to a more cosistet ad accurate positioig capability tha BDS applied stadaloe, especially at BDS-difficult area. BDS ca costrai the icreasig errors of INS, whereas INS also possesses excellet short-term precisio, which ca help mitigate the multipath effect ad clock bias of BDS (Zhag ad Xu, 2012). BDS sigals i good quality caot be always maitaied as there are BDS dark territories (like bridges, tuels ad valleys), the performace of the BDS/INS itegratio avigatio system degrades greatly because of the characteristics of the Iertial Measuremet Uit (IMU) the sesor errors would be accumulated with time whe it operates i a stadaloe mode. I order to obtai the real-time ad precise positioig of the trai, it eeds some assistat techologies to improve the itegrity ad accuracy of BDS/INS avigatio system, such as odometers, digital track maps, eddy-curret sesors, Doppler velocity sesors (Acharya et al, 2011) ad wayside traspoders like RFID or Balises (Stadlma, 2006). A commo method to overcome the GNSS obstructio is to lookig for a GNSS-alterative method, like fusig the odometer ad IMU whe i such area ad applyig a dead-reckoig priciple. However a trai positio computig from dead reckoig by odometer may fail whe worse adhesio occurs betwee trai wheel ad railway (Erest et al, 2004). Map-matchig (MM) method plays a importat role to ehace the accuracy. Based o the available track geometry ad positioig data, a rage of mathematical methods have bee utilized i the MM algorithms (Quddus ad Washigto, 2015). MM method is maily applied to idetify the curves (Saab, 2000), whe trai rus across the o-turout sectio, computig the projectig poit o track map, so as to obtai the corrected trai positio. This method is able to help esure the accuracy requiremet of real-time positioig whe successfully implemeted. O aother case, whe trai rus across the turout sectio, it is ecessary to determie the track umber at the first time with curve-fittig method. The railway map database cotais geometric coordiates which stored i discrete poits, which could cotribute to a cost cosumig of the operatio o regioal railway lies with low traffic desity (Gerlach ad Rahmig, 2009). The railway map database is usually collected from various sources, such as a GNSS had-held or extracted from aerial or satellite (Heirich, 2016). I this paper, a positioig methodology based o BDS/INS/odometer/MM itegratio is proposed, which applyig the data fusio algorithm for trai avigatio applicatios. This method is proposed to solve the problem of trai positioig durig BDS outages, the seamless trasitio of the trai operatio i various scearios ca be therefore maitaied. Whe the trai operated i the ope-sky eviromets, the BDS sigals are available to provide accurate positioig, BDS ad INS are itegrated to correct the INS errors. The INS gyroscope ad accelerometer errors are costraied by BDS measuremets ad odometer calculates velocity i avigatio frame. Meawhile, the itegratio system also delivers accurate positioig parameters i high updated rate. I the case of BDS sigals blocked, the itegrated system ca

4 be seamless switched to the INS/odometer itegratio mode, the itegrated system corrects the INS errors by usig odometer measuremets so as to provide a cotiuous ad acceptable positioig performace. Furthermore, with the shortest path search algorithm (SPSA) applied, the trai positio is projected to the digital track map to calculate the perpedicular distace, hece the correspodig 3D coordiates i the avigatio frame could be further determied precisely. I order to evaluate the proposed system, a real experimet was coducted i the wester regios of Chia. The experimetal results idicate that the proposed system ca provide the accurate ad cotiuous positioig service i both ope-sky ad BDS obstructed eviromets. This paper is structured as follows. The avigatio compoets of the proposed BDS/IMU avigatio system odometer/imu avigatio system ad MM techology are first itroduced i Sectio 2. The proposed system operatio i BDS ope-sky ad BDS-difficult scearios ad the seamless trasitio are discussed i Sectio 3. Fially, the seamless ad accurate positioig experimet is described, icludig the system cofiguratio ad experimet desig, ad the test data aalysis ad evaluatio are preseted. 2. NAVIGATION COMPONENTS AND SYSTEM DESIGN 2.1 Algorithm developmet based o BDS/INS The BDS/INS avigatio algorithm geerates positio, velocity ad attitude (PVA) solutio with loosely-coupled Exteded Kalma filter (EKF). PVA estimates are propagated forward i time based o IMU measuremets, ad errors are tracked ad compesated by utilizig a errorstate Kalma filter (Salmo ad Bevly, 2014). A. INS Update Attitude iformatio is maitaied i the form of a trasformatio matrix, relatig the trai body frame to the chose avigatio frame which is chose orth-east-dow i this paper. The subscript b,, e, i represet the body frame, avigatio frame, earth fixed frame ad iertial frame. The body coordiate frame is with the Euler rotatio agles of ( ),which meas roll, pitch, ad yaw i this paper. The iitializatio attitude matrix calculated by the three gyroscope measuremets is as follows, C b,, b C cos cos si cos si si cos cos si si cos cos si si si cos si si si cos si cos cos si si si cos cos cos (1) The attitude matrix update is implemeted usig the quaterio approach. The equatio is show as, 1 b q bq q 2 (2) b where q ad q refer to the attitude quaterio ad its rate, ad b is the skew-symmetric

5 matrix of the agular-rate. Accelerometer measuremets ca be coverted from body frame to avigatio frame ad the used i (3) ad (4) to update the positio ad velocity, P V P (3) e V f g 2 V (4) e ie where P ad P are the positio ad positio-rate vectors; V ad V are the velocity ad velocity-rate vectors; is the accelerometer measuremets; is the gravity; is the f agular rate vector of the avigatio frame with respect to the earth fixed frame, resolved i the avigatio frame ; is the agular rate vector of the earth fixed frame with respect to the iertial frame, resolved i the avigatio frame. ie B. Exteded Kalma Filter The system state i the exteded Kalma Filter is composed by eightee states icludig attitude error (roll, pitch ad yaw), positio error (latitude, logitude ad height), velocity error, gyroscopes bias, accelerometers bias ad lever arm oise of BDS, respect to IMU ceter. These state variables could be defied as, g e X P V b b l (5) g a BDS where,,, bg, ad lbds are the attitude error, positio error, velocity error, gyroscopes bias, accelerometers bias ad the lever arm oise of BDS. P V b a The Kalma filter icludes time update process ad measuremet update process, the dyamic model ad measuremet model are show i (6) ad (7), X F X (6) Z H X (7) F where is the state trasitio matrix, is the system measuremet matrix, H computed Jacobia matrix, ad are the process oise ad measuremet oise. Z is the The Kalma filter time update process is to propagate the system state X, ad the state covariace matrix P, show i (8) ad (9), where Q is the covariace matrix of the system process oise. X F X (8) T P F P F Q (9) The INS avigatio solutio calculated durig the time update is corrected i the measuremet

6 update usig the logitude, latitude, altitude wheever the BDS measuremets are available. The Kalma measuremet update process ca be performed as see i (10) (12), P H T K H P H T R (10) K X X (11) Z H X 18 P I K H P (12) The lever arm should be estimated ad compesated as the BDS atea is displaced from the IMU, ad the avigatio errors caused by the lever arm accumulate rapidly with time (Liu ad He, 2002). Therefore, a compesatio method should be take ito accout to elimiate the disturbace force by the lever arm i the itegratio avigatio system. The measuremet matrix H, is calculated as, ˆ b ˆ H3 18 Cb l0 I C b (13) where b l0 is the iitial estimate of the lever arm. System measuremet Z, is calculated by comparig the positio from BDS ad predicted positio derived from the INS, also the lever arm of BDS respect to IMU ceter should be cosidered. The Kalma filter measuremet is show as, INS BDS BDS Z P l P (14) P INS where is the INS positio solutios, positio solutios, respectively. l BDS is the estimated lever arm, ad P BDS is the BDS The covariace of the measuremet oise RBDS R BDS where the uit of measuremet oise variace is, is described as, diag (15) 2 m Algorithm developmet based o INS/odometer The use of odometer is particularly recommeded whe a LDL trai localizatio system is implemeted ad there is the risk of BDS coverage failure, which is proe to happe whe the trai eters a tuel or deep valleys (Hemerl ad Schad, 2008). The odometer is used to measure the trai wheel speed ad obtai the distace, ad works together with the INS as dead-reckoig system to determie the positio. Odometer ca provide absolute velocity as show as,

7 V odo N Q odo d (16) odo where N odo is the umber of pulses per secod couted by the odometer; Q odo is the umber of pulses whe the wheel rotates a full tur; ad d is the wheel diameter. Combied with attitude matrix, the velocity i the avigatio frame calculated by odometer parameter, which is, V C V 0 0 T (17) b odo The INS/odometer avigatio algorithm is set up i a similar maer to the BDS/INS avigatio algorithm. Where the system state has 18 compoets, icludig attitude error (roll, pitch ad yaw), positio error(latitude, logitude ad height), velocity error, gyroscopes bias, accelerometers bias ad lever arm oise of odometer, respect to IMU ceter. Hece the state vector is defied as, X P V b b l (18) g a ODO where lodo is the lever arm oise of odometer. Measuremet Z is calculated by comparig the positio from odometer ad predicted positio derived from the INS, the lever arm of odometer respect to IMU ceter should also be cosidered. The odometer positio is produced by multiplyig the velocity i the avigatio frame ad cycle time of odometer. The measuremet is show as, INS ODO ODO Z P l P (19) where P INS is the INS positio solutios; l ODO is the estimated lever arm of odometer; ad P ODO is the positio calculated by the velocity of odometer ad last trai positio iformatio. The covariace of the measuremet oise R RODO ODO where the uit of the measuremet oise variace is is described as, diag (20) 2 m Algorithm of map-matchig As the most part of railway lies are desiged as sigle track railway, the precise trai positioig by low-cost sesors ad digital map is more suitable tha covetioal trai positioig system with large umber of balises istalled alog the track. There is o doubt that the high precisio track map database is crucial to the MM method. It

8 carries lie iformatio ad plays a importat role to improve trai positioig accuracy (Saab, 2000).The digital track map is ormally geerated before the railway operatio by a series precisely measured track poits. The distace of ay two is several meters. The track poits are the processed by further optimizatio like smoothig, iterpolatio ad reductio, to geerate the formal rail track map. The positio of the trai is ormally betwee two cosecutive track poits. Based o the curvature radius of rail tracks, the trajectory betwee ay two cosecutive track poits withi the 1.5 m iterval ca be assumed to be a straight lie (Zheg ad Cross, 2012).Therefore, it is easy to obtai the trai positio by a liear mathematical model if the positios of the previous ad post track poits are kow. Fig.2 shows a example of the processig strategy. POI o track Estimated positio Track lie Map-matchig positio Figure 2. Example of MM procedure The steps are as follows, Step 1 Accordig to the estimated positio poit G( x, y ), search the off-lie track map ad calculate the earest poit from the track poit-of-iterest (POI) database. Step 2 Search the previous ad post track poits Pi ( xi, yi ) Pi 1( xi 1, yi 1) from the track POI database which guaratee the trai travels betwee them. If xi xi 1, go to Step 3; if yi yi 1, go to Step 4; otherwise jump to the Step 5. Step 3 Whe xi xi 1, the coordiates of MM positio poit is ( x, y ), go to Step 7. Step 4 Whe yi yi 1, the coordiates of MM positio poit is ( xy, ), go to Step 7. i i Step 5 Calculate the slope k betwee the two track poits, the fuctio is show as, k y x i1 i1 yi x i (21) Step 6 Get the matchig positio poit by perpedicular projectio. The fuctio is show as, 2 x ky k( yi 1 kxi 1) kx k y ( yi 1 kxi 1) M ( x, y) (, ) 2 2 k 1 k 1 (22) Step 7 If a ew poit is received, go to step 1.

9 3. SYSTEM OPERATES IN BDS-AVAILABLE AND BDS-DIFFICULT SCENARIOS The devices of trai avigatio system are istalled i the drivig cab, ad collect redudat measuremets from positioig sesors, icludig a BDS receiver, a IMU ad a odometer. The iformatio from these sesors geerate the positioig result with assistace of track map database, which is stored i the o-board map server. The, the locatio message is delivered to the mai cotrol uit. Fig.3 shows the structure of the trai avigatio system o-board equipmets. BDS satellites IMU BDS receiver Gyroscopes Accelerometer Odometer Track Map Mai Cotrol Uit Railway Track Figure 3. Structure of the trai avigatio system o-board equipmets At the begiig of the avigatio system, the BDS measuremets are available to provide accurate positioig service. The BDS/INS itegratio avigatio system aims to utilize the advatages of the two idividual systems ad overcome their weakesses. Whe BDS measuremets are available, the itegrated system ca provide a complete avigatio parameters, icludig positio, velocity, acceleratio, ad attitude at high data rates. At the same time, the velocity obtaied by odometer are combied with corrected attitude matrix ad would be stored i the mai cotrol uit. Whe trai rus close or i the sigal obstructed areas, like bridges, tuels or valleys, the quality of BDS sigal would decrease greatly. The trai positioig systems detect the BDS availability ad make the decisio to switch the system to the INS/odometer itegratio mode immediately, ad correct the INS errors by usig odometer measuremets so as to provide cotiuous ad acceptable positioig performace. The seamless trasitio of the trai operatio i differet scearios ca be therefore maitaied. The trai positios estimated via EKF the combie the off-lie digital track map data to obtai the corrected positios which should be o the railway track. The scheme to fuse the BDS, INS, odometer ad MM is show i Fig.4.

10 Railway Track Odometer BDS Odometer IMU + - BDS/INS Kalma filter ODO/INS Kalma filter - + IMU Estimated offset Estimated offset BDS is available BDS is ot available Estimated positio Track Digital Map Positio of the trai Figure 4. Trai localizatio fusio scheme There are two importat poits of the itegratio system. 1) The trasitio coditios: The umber of visible satellites ad the horizotal dilutio of precisio (HDOP) are cosidered to determie the operatio scheme of the trai positioig system, which are obtaied from the BDS measuremets. The greater the umber of visible satellites ad the smaller of HDOP, the better positioig performace could be achieved. 2) The threshold of MM: There should be a perpedicular distace threshold whe applyig the MM approach. The positio of trai is o the rail track, thus choosig a suitable threshold is the key problem. Firstly, calculate the perpedicular distace from estimated positio to the railway track, the the MM processig is implemeted whe the perpedicular distace is less tha the correspodig threshold. The proper threshold is related to the accuracy of BDS, to guaratee the effectiveess ad accuracy of trai positioig. 4. EXPERIMENT AND RESULT ANALYSIS To evaluate the performace of the proposed avigatio system, the field test was coducted i Qighai-Tibet railway i April The experimet lasted for 25 miutes ad the trai ru at

11 aroud 60 km/h. The test trajectory is show i Fig. 5 where it ru i the mout area eviromets, where BDS outages happeed frequetly. The test legth was aroud 23.5 km. The devices that were used i the test icluded oe Uicore UB370 BDS receiver, oe ADIS IMU, ad oe odometer istalled coectig to the trai wheel. The BDS positioig data ad IMU raw data are obtaied i 5Hz ad 123Hz respectively, ad the odometer output rate is 1Hz. The umber of pulses whe the wheel has rotated by a full tur is The track map data was previously geerated ad stored i the map server. Figure 5. The trajectory of the rail test coducted o Qighai-Tibet lie The o-board equipmet is istalled i the drivig cab, to obtai the INS data, odometer data, ad the BDS positioig data, the fusig the data with itegrated method ad MM techology. Fially, the trai positioig report is delivered to the mai cotrol uit. The left plot of Fig.6 is the test trai ad the o-board equipmet is show i right plot. Figure 6. The test trai ad the o-board equipmet The Fig.7 illustrates the positioig result computed by BDS, the zoom-i plot shows the trajectory o BDS-difficult period. The blue solid lie idicates the real track data of map database, while the red lie with is the BDS positioig results. It ca be see the BDS obstructed area is obvious. Fig.8 shows the offset betwee BDS positioig result ad real track, it ca be see the offset is large, that is because the trai ru i mout area ad the BDS positio was processed i the sigle poit positioig (SPP) mode with peudorage measuremets.

12 Figure 7. The positioig result computed by BDS Figure 8. The offset betwee BDS positioig result ad real track Fig.9 shows the visible satellites umber ad HDOP of BDS positioig system. Oe ca see the umber of visible satellite dropped durig the latter part of the test. At the same time, the HDOP icreased correspodigly. Figure 9. The visible satellites umber ad HDOP of BDS positioig system I order to get the better performace of trai positioig, the umber of visible satellite is more tha 6 ad HDOP is lower tha 1.5 are treated as the judgig coditios of BDS availability. Whe the umber of BDS visible satellite is less tha 6 or the HDOP is more tha 1.5, the BDS sigals are treated as uavailable, ad vice versa. The perpedicular distace threshold of

13 MM is desiged as 25m i this paper accordig to the BDS SPP accuracy. Fig.10 shows the compariso of the positioig results of BDS/MM system ad BDS positioig system. The blue solid lie idicates the real track, the red lie with is the BDS positioig results, while the gree lie with is the itegratio positioig results of BDS ad MM techology. Accordig to this figure, the MM algorithm is effective whe the system is able to receive the BDS sigals, while the itegrated system is ot available whe trai rus ito the BDS obstructed area. Figure 10. The compariso of BDS positioig ad BDS/MM positioig results Fig.11 compares the positioig performace of BDS/INS system ad BDS positioig system. The blue solid lie idicates the real track, the red lie with is the BDS positioig results, while the gree lie with is the itegrated positioig results of BDS ad INS. Whe the trai operates i the ope-sky eviromet, the BDS ad INS itegrate ad deliver the accurate positioig solutio i high updated rate. Whe the trai rus ito the tuels, oly the INS could work, the gyroscope ad accelerometer bias accumulate with time quickly, which have a serious effect for the performace of trai positioig, ad result i a uacceptable positioig solutio i very short time period. Figure 11. The compariso of BDS positioig ad BDS/INS positioig Fig.12 compares the positioig performace computed by the proposed BDS/INS/odometer/MM system ad BDS positioig system. The blue solid lie idicates the real track, the gree lie with is the BDS positioig result, while the red lie with o is the BDS/INS/odometer/MM itegratio positioig results. Whe the BDS sigals are available to provide positioig service, the BDS, INS ad MM are itegrated to obtai the itegratio positio ad at the same time, the velocity obtaied by odometer combied with estimated attitude matrix are stored i the mai cotrol uit. Whe the trai rus ito BDS-

14 obstructed area, the INS ad odometer ca still work. I that case, the odometer ad INS are itegrated with the assistat of the MM to provide the cotiuous ad seamless positioig service. The BDS/INS/odometer/MM itegratio positioig result shows the itegrated system ca be seamless switched to the INS/odometer/MM itegratio mode i the case of BDS sigals blocked ad the matchig poits are i good agreemet with the real railway track. Figure 12. The compariso of BDS/INS/odometer/MM positioig ad BDS positioig 5. CONCLUSIONS This paper proposed a itegrated avigatio system based o BDS, INS, odometer ad MM techologies i order to provide the cotiuous ad reliable trai positioig service i LDL railway. The proposed avigatio system has two operatig scearios: ope-sky ad BDS obstructed eviromets. I the ope-sky eviromet, the BDS sigals are available to provide accurate trai positioig, the BDS ad INS are itegrated to correct the INS sesor errors via EKF. O the other had, the avigatio system switch to the INS/odometer itegratio mode whe the trai rus close or eters a BDS-difficult area, like tuel, valley, bridges or hillside, to esure the reliability ad accuracy of trai real-time avigatio. Furthermore, the algorithm of MM ca sigificatly reduce the excessive errors of BDS SPP solutio ad help determie the trai locatio o rail track. To evaluate the proposed avigatio system, a real experimet was coducted o Qighai-Tibet railway i the wester regio of Chia. The experimetal results cofirm that the proposed system ca provide the seamless ad cotiuous positioig service i both ope-sky ad BDS obstructed eviromets. Compared with a covetioal BDS ad BDS/INS avigatio system, the proposed system shows improved system performace i differet scearios, especially o the BDS-difficult area. ACKNOWLEDGEMENTS This work was supported i part by the State Key Program of Natioal Natural Sciece

15 Foudatio of Chia uder Grat ad U , the Natioal Iteratioal Scietific ad Techological Cooperatio Project uder Grat 2014DFA80260, ad the Fudametal Research program for Cetral Uiversities uder Grat 2016RC020. REFERENCES Soderholm S, Bhuiya M, Thombre S, Ruotsalaie L, Kuusiemi H (2016) A multi-gnss softwaredefied receiver: desig, implemetatio, ad performace beefits, Aals of Telecommuicatios 71(7-8):1-12 Zhag T, Xu X (2012) A ew method of seamless lad avigatio for GPS/INS itegrated system, Measuremet 45(4): Acharya A, Smita S, Ghoshal T (2011) Trai localizatio ad partig detectio usig data fusio, Trasportatio Research Part C 19(1):75-84 Stadlma B (2006) Automatio of operatioal trai cotrol o regioal brach lies by a basic trai cotrol, Proc. IEEE Itelliget Trasportatio Systems Coferece, Toroto, Caada, September, Erest P, Mazl R, Preucil L (2004) Trai locator usig iertial sesors ad odometer, Proc. IEEE Itelliget Vehicles Symposium, Parma, Italy,14-17 Jue, Quddus M, Washigto S (2015) Shortest path ad vehicle trajectory aided map-matchig for low frequecy GPS data, Trasportatio Research Part C 55: Saab S (2000) A map matchig approach for trai positioig part II: applicatio ad experimetatio, IEEE Trasactios o Vehicular Techology, 49(2): Gerlach K, Rahmig C (2009) Multi-hypothesis based map-matchig algorithm for precise trai positioig, 12th Iteratioal Coferece o Iformatio Fusio, Seattle, Washigto, USA, 6-9 July, Heirich O (2016) Bayesia trai localizatio with particle filter, loosely coupled GNSS, IMU, ad a track map, Joural of Sesors, 2016:1-15 Salmo D, Bevly D (2014) A exploratio of low-cost sesor ad vehicle model Solutios for groud vehicle avigatio, 2014 IEEE/ION Positio, Locatio ad Navigatio Symposium (PLANS 2014), Moterey, Califoria, USA, 5-8 May, Liu J, He X (2002) Aalysis of lever arm effects i GPS/IMU itegratio system, Trasactios of Najig Uiversity of Aeroautics & Astroautics 19(1):59-64 Hemerly E, Schad V (2008) Implemetatio of a GPS/INS/ODOMETER Navigatio System, ABCM Symposium Series i Mechatroics 3: Saab S (2000) A Map Matchig Approach for Trai Positioig Part I: Developmet ad Aalysis, IEEE Trasactios o Vehicular Techology 49(2): Zheg Y, Cross P (2012) Itegrated GNSS with differet accuracy of track database for safety-critical railway cotrol systems, GPS Solutios 16(2):