You are on page 1of 12

Available online at www.sciencedirect.

com

ScienceDirect
Advances in Space Research 55 (2015) 113–124
www.elsevier.com/locate/asr

BDS/GPS relative positioning for long baseline with
undifferenced observations
Min Wang ⇑, Hongzhou Cai, Zongpeng Pan
Zhengzhou Institute of Surveying and Mapping, No. 62 Kexue Road, Zhengzhou 450001, China
Received 22 July 2014; received in revised form 30 September 2014; accepted 1 October 2014
Available online 13 October 2014

Abstract
Before and after the official beginning of Beidou navigation satellite system (BDS) regional service on December 27, 2012, many applications based on BDS such as real-time kinematic (RTK) and precise point positioning (PPP) with real data have been considered in the
literatures. However, lack of precise satellite antenna correction and relatively low quality of BDS orbit and clock product is an obstacle
for PPP and relative positioning over long baseline using BDS observations. In this paper, the Double Station Observation Processing
(DSOP) method that directly uses undifferenced data is applied to relative positioning. By estimating the satellite clock offsets on-the-fly,
the satellite dependent unmodelled error can be compensated. Moreover, the direct use of undifferenced observation makes the method
easy to implement and flexible to adapt observations of multiple systems. Experiment results demonstrate that relative positioning based
on BDS observations can be achieved at centimeter accuracy level which is better than conventional PPP results with limited computation burden increase. These results also indicate the promising potential of BDS to develop real-time service.
Ó 2014 COSPAR. Published by Elsevier Ltd. All rights reserved.

Keywords: BDS/GPS; Long baseline; Undifferenced observation; Real-time; Orbit prediction

1. Introduction
Chinese Beidou navigation satellite system (BDS) officially began to provide regional service for Asia–pacific
users on December 27, 2012. The operational constellation
of BDS includes 14 satellites, of which the satellites of C01–
C05 are in geostationary orbit (GEO), the satellites of C06–
C10 are in inclined geosynchronous orbit (IGSO) and the
satellites of C11–C14 are in Medium Earth orbit (MEO).
Several approaches about the positioning algorithms and
applications of BDS have been considered in the literatures. Currently, BDS can provide applicable real-time
kinematic (RTK) positioning over baseline shorter than
20 km for Asia–Pacific user. Combing the observations of
GPS with BDS, the RTK results are more reliable than
⇑ Corresponding author. Tel.: +86 15003990576.

E-mail address: difzibo@163.com (M. Wang).
http://dx.doi.org/10.1016/j.asr.2014.10.001
0273-1177/Ó 2014 COSPAR. Published by Elsevier Ltd. All rights reserved.

one system only (Deng et al., 2014; He et al., 2014; Shi
et al., 2013; Teunissen et al., 2014). Furthermore, all BDS
satellites broadcast signal in three frequencies, which
makes the assessment and study of triple-frequency positioning technique more feasible. Experiment results with
real data of BDS show that the success rate of the ambiguity fixing of triple-frequency is higher than the dual-frequency case for baseline shorter than 43 km, which is
beneficial to improve the positioning reliability (Li et al.,
2013a; Tang et al., 2014). Besides, an important benefit
from triple-frequency GNSS is the enhanced capability to
detect phase slip (Teunissen and Bakker, 2013).
The results of BDS Precise Orbit Determination (POD)
are reported by scientists in China and abroad
(Montenbruck et al., 2013; Shi et al., 2013; Steigenberger
et al., 2013; Zhao et al., 2013). Current POD method for
BDS is essentially the same as that used for GPS POD.
Limited by the size of tracking station network and lack

1. GNSS pseudorange and phase observation model can be presented as follows (Jin et al. and Lji. so the BDS results need more time to converge compared to GPS and GLONASS ones (Li and Zhang. Moreover. / Advances in Space Research 55 (2015) 113–124 of knowledge of satellite (such as phase center offset (PCO). Observation model Without loss of simplicity. 2014).f þ ei. which is very important and helpful resource for the research of BDS precise positioning (http://www.igs.. we limit POP to two stations observation processing to handle the relative positioning application and both BDS and GPS observations are introduced in relative positioning.f þ eji. PPP technique uses undifferenced phase and pseudorange observation for high accuracy positioning without need for dense reference stations and communication links (Kouba and He´roux. This paper is organized as follows: the methodology of DSOP is introduced first. began to submit their BDS precise orbit and clock to IGS-MGEX archives since March 23.114 M. As a result. the result for baseline longer than 1000 km is rarely reported. the lack of clear antenna correction and relatively low quality of orbit make above preconditions for double-differenced relative positioning over long baseline no longer stand for BDS. At theory level. They found that BDS could support real-time precise point positioning at decimeter or even centimeter accuracy level before the full constellation deployment was finished. Because satellite clock offset is estimated on-the-fly. which extended the concept of PPP by estimating satellite clock on-the-fly on a local station network base. Li et al. The direct adaption of undifferenced observation can avoid differencing procedures between observations of different stations and brings high flexibility for different applications. Currently.. 1997). In this paper. (2010) introduced a “Precise Orbits Positioning” (POP) technique. dtsys is the system time difference between GPS and other systems. 2. we apply the DOSP for the GPS/BDS relative positioning which is inspired by the POP method introduced by Salazar et al. the solution is sensitive to orbit error and accuracy of PCO and PCV correction. the unmodelled satellite dependent errors e. With high quality orbit and clock products and accurate PCO and PCV correction. 2008. Methodology The application of undifferenced observation for data processing is not a new idea. 2009). centimeter and millimeter accuracy level relative positioning for long baseline can be achieved with only BDS observations. as one of IGS-MGEX (Multi-GNSS Global Experiment) analysis center (Montenbruck et al. (2013b) experimentally adapted BDS into GFZ real-time service that operates as a part of IGS real-time service. Due to weak geometry of GEO satellites. 2001. Salazar et al. However. 2014). the subscript i and f denote the receiver and signal frequency respectively. followed by the assessment experiments. 2010):   P ji.f ð1Þ   j j j j j Li. Wang et al. user station coordinates are estimated under post-processing and real-time environments.org/mgex/ accessed on 2014/04/05). success rate of BDS ambiguity fixing is lower than GPS and little improvement of positioning accuracy for BDS is achieved even after ambiguity fixing procedure (Feng et al. At user end. some conclusions and remarks are given. The ambiguity is left as float value and the satellite clock offsets are estimated on-the-fly to absorb some satellite dependent errors (such as orbit error and incomplete PCO correction) that cannot be fully modeled.f where both Eqs. As to BDS relative positioning.f þ kf N i. the GEO orbit accuracy is the worst among the three types of satellites comprising BDS spatial segment. 2. a Double Station Observation Processing (DSOP hereinafter) method is introduced that enables precise relative positioning for baseline up to 2551 km. Zumberge et al. This technique yields a higher positioning rate and shorter convergence time compared to PPP technique.f ¼ qi þ c dti  dtj þ dtsys þ MdT i  I i. Wuhan University. Xu (2007) proves the equivalence of the differenced and undifferenced data processing method.g. (1) and (2) are in units of distance. 2013). the direct use of undifferenced observation makes this method easy to adapt observations of multiple systems and suitable for real-time application. respectively. BDS orbit is not as precise as GPS orbit delivered as part of IGS (International GNSS Service) final product.f ð2Þ P ji. dtj is the satellite clock offset. c is the light speed in vacuum. the superscript j denotes the satellite. 2014. qji is the geometry distance from receiver to satellite.. In this paper. conventional double-differenced process can eliminate or decrease the satellite and receiver dependent errors in GPS observations. Different from POP method. For PPP technique. this term exists only when the equation does not belong to the GPS .f are the pseudorange and phase observation respectively. Finally. satellite antenna phase center variation and orbit error can be compensated. which is beneficial to exploit integer nature of phase ambiguity and makes the result more stable and accurate.. Several GNSS applications at network end like POD and high accuracy satellite clock estimation directly use undifferenced observation (Laurichesse et al.. By estimating satellite clock offset on-the-fly. which makes relative positioning under real-time environment possible. unclear knowledge of satellite antenna and relatively low quality orbit are limitations for relative positioning over long baseline and precise point positioning (PPP) based on BDS. Therefore. dti is the receiver clock offset. In the experiments. only predicted orbit and satellite clock derived from broadcast ephemeris are used. phase center variation (PCV) and solar radiation pressure (SRP) model).f ¼ qji þ c dti  dtj þ dtsys þ MdT i þ I ji. (2010).

eji. weather monitoring and time synchronization.2. receiver and satellite clock offset. different from double differenced observation based technique. the observation models for reference station and user station can be organized as follows: Reference station: ljB. the subscript B and R represent the reference station and user station respectively.LC and Lji. kLC and 1 2 1 2 N j.LC are the pseudorange and phase ionosphere-free observation respectively. (5) and (6) can be cancelled out: ljB.2 ¼ qji þ c dti  dtj þ dtsys þ MdT i þ eji. 2014):   P ji. 2009).M. Different methods are used to model the stochastic characteristic of unknown states.LC ¼ cdtB  cdtj þ dtsys þ MdT B þ kLC N jB. y j and zj are the satellite coordinates for satellite j.f contain noise and other unmodelled errors for phase and pseudorange observation.LC ¼ aP ji.LC ljR. Two stations are involved in data processing of relative positioning. (7) and (8) are left for estimation. the user station coordinates are modeled as constant state. b ¼ f 2 f2 2 .f is the phase ambiguity that is not integer since it includes the receiver and satellite initial phase biases. 2010). this equation system can be easily apply to relative positioning with observations of multiple systems. Kalman filter is used for unknown state estimation in this paper.f and eji. Before the coordinate estimation procedure. respectively. 1996). which enables real-time application of this method.LC ð3Þ   j Lji.PC ¼ cdtB  cdtj þ dtsys þ MdT B ð6Þ User station: ljR.2 ¼ qji þ c dti  dt þ dtsys þ MdT i þ kLC N ji. the troposphere delay and receiver clock parameters are preserved. I ji is the slant ionosphere delay. We only estimate the rest part of ZWD values actually. so the transition matrix for clock offset is zero matrix. observation noise and unmodelled error items are ignored. the unknown state can be grouped as user coordinate. M is the troposphere mapping function. one external constrain must be add to the equation system.LC represents the observation minus computation (OMC) terms for phase observation.LC ð4Þ where P ji. 2. Since ZHD values and prior ZWD values can be corrected from observation with empirical troposphere models like Saastamoinen model. Wang et al. the receiver clock items in Eqs. a and b are the f2 f 2 combination coefficients and a ¼ f 2 f1 2 . For relative positioning of the static station.LC ð5Þ ljB. The receiver clock of reference station is forced to be zero as reference clock. because only prior values of satellite clock offsets are needed. The user station coordinates are left for estimation and reference station coordinates are fixed to known values therefore no longer exist in the reference station equation. the resulting 115 normal equation matrix exhibits rank deficiency of size one.LC ¼ aLji. The Niell mapping functions are used to convert the zenith troposphere delay to slant direction (Niell. T i is the zenith troposphere delay (ZTD). and they are named as user station and reference station. real-valued ambiguity. (7)–(10). and all other clock offsets are estimated respect to it.LC are the wavelength and phase ambiguity for ionosphere-free phase observation respectively. y 0 and z0 are the prior coordinates for user station.1 þ bP ji. xj . / Advances in Space Research 55 (2015) 113–124 observations. which means that the clock offset of every epoch is treated as independent time-discrete state. the satellite clocks in broadcast ephemeris are accurate enough. zenith troposphere delay and system time difference. all the observation corrections are performed as those used for conventional PPP technique. The ionosphere-free combination is used to cancel out the ionosphere error (Jin and Najibi.LC ljB. Hence. The process-noise matrix for clock offsets are diagonal matrix and the diagonal values of them are set to a relatively large value to adapt high temporal variation characteristic of clock offset. The connection between reference and user station is achieved by simultaneous estimation of satellite clock offsets (Salazar. Furthermore. The zenith troposphere delay contains Zenith Hydrostatic Delay (ZHD) and Zenith Wet Delay (ZWD). Stochastic model In the observation equation system described above. The GPT model can be used to calculate the local temperature and pressure and then feed Saastamoinen model to calculate ZHD and prior ZWD values (Kouba. which is particularly important for BDS relative positioning. White noise model is used to model the receiver and satellite clock offset. The satellite clock offset can also absorb some unmodelled errors at satellite end.LC ¼ cdtj þ dtsys þ MdT B þ kLC N jB. To solve this problem.g.PC ¼ ! x0  xj dx þ qj0 ! y0  yj dy þ qj0 ! z0  zj dz qj0 þ cdtR  cdtj þ dtsys þ MdT R þ kLC N jR. lj. On the other hand.PC ¼ cdtj þ dtsys þ MdT B ð9Þ ð10Þ Because only undifferenced observations are involved in Eqs.LC þ eji.LC ! ! ! x0  xj y0  yj z0  zj ¼ dx þ dy þ dz qj0 qj0 qj0 þ cdtR  cdtj þ dtsys þ MdT R ð7Þ ð8Þ where lj. If both the receiver clock and satellite clock parameters in Eqs. The transition matrix of station coordinates is an identity matrix and their process- . kf is the wavelength of phase observation. which can be useful information for other application purposes e.PC represents the OMC terms for pseudorange observation.1 þ bLji. N ji. After the observation combination and linearization process. x0 . respectively.

The IGS ultra-rapid (IGU) product has already served as a source of GPS orbit for real-time or near real-time application for more than 10 years (Dow et al.S are nine parameters of solar pressure model. Then the resulted parameters are used to predict future orbit. Ambiguity state is modeled as constant state when the satellite is continuously tracked. Fig. The BDS precise orbit of day of year (DOY) 86–99 generated by Wuhan University are used to predict the BDS orbit. The IGU product is released four times a day with a latency of 3 h at 03:00 UTC.S . (11): Qstate ¼ qstate  Dt ð11Þ where qstate is the spectral density for estimating state and Dt is the estimation sample rate. such routine predicted product is not currently available.C. 1). DðuÞ. r is the length of geocentric radius vector r of the satellite. The satellite movement acceleration can be described by the so-called equations of motion (Dach et al. .Þ ð12Þ r where GM is the product of the constant of gravity and the mass of the earth. a is the perturbation acceleration characterized by many parameters. Therefore.C. . 3. 15:00 UTC and 21:00 UTC. longer the orbit arc is needed. Orientation axes of the spacecraft-centered reference frame used for solar radiation pressure effects model. 3). not all BDS satellites orbit are available in everyday IGS-MGEX product. p2. After that. The orbit accuracy of predicted half IGU product is about 5 cm. Precise orbits are used as pseudo observations to estimate six orbital elements and up to nine solar radiation pressure (SRP) parameters that describe the satellite movement.C. The 3–9 h of predicted orbits are compared with observed precise orbit to leave 3 h latency for simulating routine real-time service operation.S . which is comparable to IGS rapid (IGR) product. Wang et al. Y ðuÞ and BðuÞ are three solar radiation pressure along three axes defined in spacecraft-centered reference frame (Fig. Each IGU product is composed of 24 h of observed part and 24 h of predicted part. (13): 2 3 2 3 2 3 2 3 DðuÞ D0 DC DS 6 7 6 7 6 7 6 7 4 Y ðuÞ 5 ¼ 4 Y 0 5 þ 4 Y C 5  cosðuÞ þ 4 Y S 5  sinðuÞ BðuÞ B0 BC BS ð13Þ where u is the argument of latitude for the satellite. but the precise orbit is still needed for real-time application. 2007): r €r ¼ GM 3 þ aðt. satellite orbit is smooth and easy to predict with high accuracy. The ZWD values and system time difference are estimated as random walk (RW) process. As for BDS. Y 0. these parameters are used to predict orbit with satellite celestial mechanics knowledge. the corresponding ambiguity is initialized..116 M. 09:00 UTC. 2. The transition matrices for ZWD and system time difference are also diagonal matrices and the corresponding processnoise matrices Qstate is determined with Eq. Detailed stochastic model settings used in this paper are listed in Table 1. unlike the poor long period stability of satellite clock. In this section. p0. r. Fortunately. BDS orbit prediction The DSOP proposed in this paper decreases the accuracy need of satellite clock by estimating it on-the-fly. r_ . Unknown state Transition coefficient Process-noise ZWD System time difference Receiver/satellite clock offset Ambiguity 1 1 0 1 (continuously tracking)/0 (cycle slip or beginning of observation) 1 2.. 1. D0. In the case of orbit prediction. the GM rr3 is the two-body term of gravity field. C 0. the precise orbits are used to predict for real-time use. Since the precise orbit and clock of BDS are not released with earth rotation . p1. The solar radiation pressure term is characterized as Eq. only the solar radiation pressure is considered for perturbation influence. 2009). As presented in Fig. / Advances in Space Research 55 (2015) 113–124 Table 1 Stochastic model settings for different unknown states. Precise orbits are re-expressed and fitted in a form of orbit elements and perturbation parameters. we use the longest orbit arc as possible in past 72 h precise orbit files to predict 9 h orbit of every satellite with six orbital parameters and nine SRP parameters scenario (as depicted in Fig. harder to find a consecutive arc for orbit prediction. . When cycle slip happens or the satellite rises.7e7 m2/s  Dt 1e7 m2/s  Dt 1e10 m2 0 (continuously tracking)/4e8 m2 (cycle slip or beginning of observation) 0 User station coordinate (static station) noise is set as zero.

Experiments and result analysis 4. Fig. the convergence speeds of the solutions under different environments are discussed. The weekly average solutions of conventional PPP technique acquired with PANDA software package developed by GNSS Research Center. Then predicted BDS orbit and predicted part of IGU product are used to assess the performance under real-time environment. All observations are processed on a day by day basis. 4.M. four BETN stations (BJF1. parameters (ERPs).1. The PCO and PCV correction information of GPS satellite antenna generated and released by IGS in the form of ‘antex’ file is used for GPS observation.1 km. cross-tracking and radial component of predicted orbits compared to precise orbit for each satellite. It can be seen that the orbit prediction accuracy evidently varies with the types of satellites.86 cm. Figs.igs. Wuhan University (Shi et al. respectively.. (2013). four baselines are composed with the baseline lengths varying from 1049. The IGS-MGEX recommended values of satellite PCO correction is applied for BDS observation (http://www. IGU predicted ERPs are used instead. the along-tracking and cross-tracking direction differences of rest satellites are comparable to each other and inherent with the POD accuracy reported in Zhao et al. All the stations of BETN are equipped with UR240 GPS/BDS receivers manufactured by Unicore Communications Incorporation. 4 and 5 present the averaged mean difference and RMS values in along-tracking. the PCO and PCV correction of receiver antenna are ignored for both . 2008) using GPS observation and IGS final products are used as external reference values for station coordinates.htm accessed on 2014/04/05). the quality of predicted GEO orbit is the worst among the three types of satellites. HRBN. The average RMS of radial direction for IGSO satellite and MEO satellite are 7. Schematic diagram of BDS orbit prediction. Mainly due to relatively low accuracy of GEO satellite (C01–C05) both as observation for prediction and comparison values. As shown in Fig. Because positioning result is less sensitive to orbit errors in along-tracking and cross-tracking direction than in radial direction. Coordinates of user stations are firstly estimated with BDS precise orbit generated by Wuhan University and IGS final product to assess the positioning accuracy in post-processing mode. In our experiments.org/mgex/Status_BDS. 2012). Wuhan University (Shi et al. The sample rate of observation is 30 s.39 cm and 2. It should be noticed that MEO and IGSO satellite orbits of BDS can be predicted at centimeter accuracy level in radial direction. And no clear relationship of orbit differences between different directions is noticed. IGSO and MEO satellites may play a more important role in high accuracy positioning for users. Due to lack of public correction values for receiver antenna. 3. Except for GEO satellites. The predicted orbit results of six hours are merged for further positioning processing. / Advances in Space Research 55 (2015) 113–124 117 Fig. CHDU and HKTU) located in China are used and the observation period is from DOY 91–98 in 2013. Wang et al.3 km to 2551. Finally. 6.. Data description and processing strategy All the BDS observations used in this paper are collected by Beidou Experimental Tracking Network (BETN) established by the GNSS Research Center. Availability of BDS precise orbit delivered by Wuhan University. 2.

2 -0. which cause a relatively large fluctuation of PDOP value in 91 day. Wang et al.8 0.4. The differences RMS of the predicted BDS orbits compared to the precise observation ones.118 M.7.2 1 0.5 for GPS. / Advances in Space Research 55 (2015) 113–124 0. 4. Post-processing results For post-processing experiment. RTK.2.6 0. GPAK is a scientific GNSS data processing software developed by Zhengzhou Institute of Surveying and Mapping and has been tested by several scientific research and development programs. 4. so the positioning results of BDS over short baseline are more .1 0 -0. The correlation of atmospheric parameters of reference station and user station diminishes as the length of baseline increase. 9 and 10. after the completion of BDS constellation deployment for regional service. the number of GPS satellite in 91 day is less than other days. 8. As presented in Figs. 2 Radial Along Cross RMS of the orbit differences (m) 1. GPS and BDS observations.7 and 16. and corresponding average number of visible satellite is 7. 3. 5. The mean differences and repeatabilities of the user station coordinates are showed in Figs.4 0.0 and 1.6 1. precise estimation of satellite clock and GNSS based atmosphere monitoring.2 0 C01 C02 C03 C04 C05 C06 C07 C08 C09 C10 C11 C12 C13 C14 Satellites Fig.2 0.3 C01 C02 C03 C04 C05 C06 C07 C08 C09 C10 C11 C12 C13 C14 Satellites Fig. For CHDU station.8 1.5 Radial Along Cross Mean of the orbit differences (m) 0.4.3 0. The mean differences of the predicted BDS orbits compared to the precise observation ones. All the observation processing including DSOP and PPP presented below are performed by the software GPAK (GNSS Processing and Analysis Kit). BDS precise orbits and clocks submitted to IGS-MGEX archive by Wuhan University and IGS final product are used. 7 and 8. the average PDOP value is 2.4 1. respectively. Its main functions include PPP.4 0. BDS and combined BDS/GPS respectively.1 -0. the satellite visibility and geometry of BDS in China are comparable to those of GPS. Due to internal tracking problem of receiver.

Whereas. PPP technique fixes the precise orbits and clock as reference and some unmodelled errors like orbit error are directly reflected in PPP solution.1km 1514. accurate than the ones for long baseline. but the current daily results of combined BDS/GPS should not be expected to be better than GPS based results. the maximum 3D mean difference is 2. Real-time results For real-time experiment. respectively. The observation preprocessing and error correction strategy applied in PPP processing is the same used for DSOP. which confirms that the relatively large satellite dependent error can be absorbed by the estimated clock offset of BDS satellite. GPS orbits and clocks from predicted part of IGU product. and 22% for baseline CHDU–HKTU. similar correlation relationship of baseline length and coordinate accuracy is not obvious for GPS results. The minimum of 3D mean difference is 0. Therefore. .57 cm for baseline BJF1–HRBN. For current BDS positioning applications where a remote reference station is available. GPS results are more accurate than the BDS ones. Wang et al. 7. possible reason is that the strength of equation system can be augmented by adding one more station to it and the on-the-fly estimated clock offset could absorb some satellite dependent unmodelled errors. The positioning results of user station HRBN are then compared with those estimated by PPP technique (Table 2). 31% for baseline CHDU–BJF1. 35% for baseline CHDU–HKTU. GPS and BDS + GPS for station CHDU.84 cm for baseline BJF1–HRBN and the maximum 3D baseline repeatability is 2. Limited by the size of tracking stations. PDOP values of BDS. Because satellite phase center are well determined and orbit quality is better. there is relatively small room for estimating satellite clock to absorb satellite dependent unmodelled errors compared to BDS case. current BDS orbit and clock quality is worse than that of IGS product.3. DSOP can provide better position estimation accuracy with limited computation burden increase compared to PPP technique.4km 30 ° N CHDU 1368. Since the quality of GPS orbit is good enough and the satellite PCO and PCV are precisely corrected. 4. especially for BDS GEO satellite. Adding GPS observations can improve the accuracy of BDS results. the PPP results accuracy is at centimeter level for horizontal direction and over 10 cm for up direction.3km 40 ° N BJF1 2551. User coordinates estimated with DSOP are one order of magnitude more precise than PPP solution. The 3D repeatability of GPS results is only 41% of BDS ones for baseline BJF1–HRBN.17 cm for baseline CHDU–HRBN. we derive ERPs. / Advances in Space Research 55 (2015) 113–124 50 ° N HRBN 1049. The minimum 3D baseline repeatability is 1. Baselines and stations processed from the test dataset.9km HKTU 20 ° N 100 ° E 110 ° E ° 130 E 120° E Fig. BDS orbits are predicted with six orbital elements and nine SRPs scenario as presented in Section 3 and prior values 7 BDS GPS BDS+GPS 6 PDOP value 5 4 3 2 1 91 92 93 94 95 96 97 98 99 Day of year 2013 Fig. 6.M. 119 Therefore.90 cm for baseline CHDU– HKTU.

78 cm respectively and the corresponding 3D baseline repeatability are 2.65 cm and 1. However. these positioning results are promising. Considering relatively low quality orbit we used for BDS. the BDS positioning results with centimeter accuracy level are still obtained in real-time.78 cm. they are precise enough for DSOP. The mean position differences of user station for four baselines under post-processing environment.01 -0. 2. which is even better than the results of post-processing PPP showed in Table 2. 0. Real-time results are presented in Figs. 1. 11 and 12. Whereas. so estimating satellite clock offset on-the-fly is more important for strengthening the equation system in real-time case.01 0. The relationship between baseline length and BDS only estimation accuracy is more evident in real-time mode than post-processing mode. the results could demonstrate the real-time performance of DSOP. BDS real-time results are worse than postprocessing ones.005 0 -0. The BDS only 3D coordinate differences of all four baselines (in order of BJF1–HRBN CHDU– BJF1 CHDU–HKTU and CHDU–HRBN) are 1. It is confirmed that even though the accuracy of predicted IGU clocks is at nanosecond accuracy level (Dow et al. 2009).37 cm and 4. 9. Mean of the position differences (m) 0. GPS and BDS + GPS for station CHDU.015 BDS N BDS E BDS U GPS N GPS E GPS U BDS+GPS N BDS+GPS E BDS+GPS U 0. of BDS satellite clocks are derived from broadcast ephemerides.015 -0.120 M. / Advances in Space Research 55 (2015) 113–124 22 BDS GPS BDS+GPS 20 Number of satellites 18 16 14 12 10 8 6 4 91 92 93 94 95 96 97 98 99 Day of year 2013 Fig.07 cm.16 cm. Wang et al. Number of visible satellites of BDS. Although the data processing is not finished under actual real-time computation environment.22 cm. All the observations are organized and processed epoch by epoch for simulating real-time one-way data stream.02 BJF1-HRBN CHDU-BJF1 CHDU-HKTU CHDU-HRBN Baseline name Fig. The GPS only results do not undermine much as the orbit source alternates from IGS final product to IGU predicted product.005 -0. 8. ..07 cm. 3.

23 3.91 1. The convergence speeds of DSOP solutions under different computation environments are presented and analyzed here.16 1.02 0.025 Coordinate repeatability (m) 121 0.39 1.70 2.03 0.70 0.45 0.44 1.37 2.01 0 -0.52 2.005 0 BJF1-HRBN CHDU-BJF1 CHDU-HKTU CHDU-HRBN Baseline name Fig. / Advances in Space Research 55 (2015) 113–124 0.14 19.58 0.04 0. but worse than the ones based on GPS.15 0.36 2. combining GPS and BDS observation is a feasible way to improve the accuracy of relative positioning using only BDS observation.19 0.04 BDS N BDS E BDS U GPS N GPS E GPS U BDS+GPS N BDS+GPS E BDS+GPS U 0. Convergence speed comparison The convergence speed is also an important index especially for the time critical applications.52 2.58 1. Wang et al.14 1. 13 shows the solution convergence for baseline CHDU– BJF1 in DOY 94.27 0. Currently. Fig.02 0.66 0. The mean position differences of user station for four baselines under real-time environment. Table 2 Comparison of BDS only user coordinate differences estimated with PPP technique and DSOP (unit: cm). the positioning Mean of the position differences (m) 0. 2013.14 0.88 0.19 2.27 0.39 1.4.34 3.01 -0.89 1. .22 8.06 16.22 0.54 0.30 0.24 4. DOY 91 92 93 94 95 96 97 98 Mean Repeatability PPP DSOP N E U N E U 0.04 3.24 0.02 BJF1-HRBN CHDU-BJF1 CHDU-HKTU CHDU-HRBN Baseline name Fig.70 0.03 BDS N BDS E BDS U GPS N GPS E GPS U BDS+GPS N BDS+GPS E BDS+GPS U 0.19 12. The combined BDS/GPS results are better than the results based on BDS. 4.63 1.88 1.01 0.51 1.36 0.67 The bold values indicate the best result of DSOP during 8 days experiment.03 0.50 12. The user station coordinates repeatabilities of four baselines under post-processing environment. 10.M. Because of relatively large orbit error and slow changing satellite geometry.35 0.56 12.36 2.34 0.97 0. 11.46 0.89 2.17 0.02 0.17 0.015 0.26 4.69 4.58 2.04 1.89 1.

2013.122 M. Comparison of real-time mode solution convergence speed of different systems for baseline CHDU-BJF1 in DOY 94. solution of BDS needs more time to converge. the addition of BDS observation to GPS observation can speed up the convergence. other combined GPS/BDS solutions are better than the results based on single system. 13 with Fig. 12.35 0. the position solution based on BDS needs more time to reach convergence in real-time mode than in post-processing mode.45 3D position error (m) 0. Except for baseline BJF1–HRBN. 2013).5 BDS GPS BDS+GPS 0.03 0. the mean 3D biases of all four baselines are less than 10 cm. the GPS and combined BDS/GPS solutions can reach the same accuracy level in less than 1 h.01 0. 0.02 0. It is worth noticing that. It is noticed that the 3D position bias of BDS needs more than 5 h to reach centimeter accuracy level in real-time. In addition.05 0 00 02 04 06 08 10 12 14 16 18 20 22 00 Hour of day Fig. 14. The user station coordinates repeatability of four baselines under real-time environment.25 0.005 0 BJF1-HRBN CHDU-BJF1 CHDU-HKTU CHDU-HRBN Baseline name Fig. / Advances in Space Research 55 (2015) 113–124 0. Like the role played by GLONASS observation in the PPP result based on combined GPS/GLONASS (Cai and Gao.025 0.1 0. combined GPS/ BDS results are more stable both in real-time mode and post-processing mode than one system only solution. the bias jump of BDS result caused by data problem at 3:45 is eliminated by adding GPS to the estimation. But the source for the orbits causes little influence on the GPS and combined BDS/GPS solutions compared with BDS solution.3 0. But GPS only solution are much accurate than BDS only ones. but the daily position accuracy may not be improved.015 0.2 0. where the solution of combined GPS/BDS is slightly worse than GPS only solution. As showed in Table 3.4 0. Comparing Fig.15 0. 13. Meanwhile.035 BDS N BDS E BDS U GPS N GPS E GPS U BDS+GPS N BDS+GPS E BDS+GPS U Coordinate repeatability (m) 0. Wang et al. .

Acknowledgements This work is supported by China National Natural Science Foundation of China (Nos.90 0. R. GNSS centre of Wuhan University is appreciated for providing BDS observation and product.26 0. A relative static positioning method that directly uses undifferenced observation is introduced.70 3.. 2013.43 2.21 1. University of Bern. the convergence speed can be improved by using combined BDS/GPS observations. DSOP can be seen as the simplest form of BDS real-time service that is composed with a reference network of only one station and one user station.79 1. Feng.07 0.. IGSMGEX project is grateful for its valuable GNSS research resource. P. 2009.02 1. Zhou. / Advances in Space Research 55 (2015) 113–124 123 0.58 0. 375–386. Gao. satellite phase center parameters of BDS have not been published and quality of experimental IGS-MGEX BDS orbit is worse than other routine IGS product. 41374041 and 41374042).E.02 0.. The assessment of DSOP is a good capability indication of BDS to develop real-time service. Fridez.. Neilan.3 0. 83...1 0.51 6.M. Meindl.5 BDS GPS BDS+GPS 0.35 0.60 0. which confines precise positioning using BDS observations. Liu. 14.12 1.53 1. References Cai. Table 3 Comparison of position difference after four hours for GPS. Wu. Bernese GPS Software Version 5. better quality of IGS product makes the GPS solutions more precise than BDS ones. Urs. Modeling and assessment of combined GPS/ GLONASS precise point positioning. J. .63 0. 2014... Rizos.51 0.53 1.56 1.61 0. Geod.47 0. M...64 0. R. the unmodelled satellite dependent errors can be compensated.44 0.90 0. Tang..15 0.. Dow. Deng. 191–198. J. Thanks also go to two reviewers for their valuable comments which have improved this paper considerably.25 0..05 1. With the development of better orbit and orbit prediction strategy. Comparison of post-processing mode solution convergence of different systems for baseline CHDU-BJF1 in DOY 94. By estimating the satellite clock offset on-the-fly. 223–236. In: Sun. However. Reliable single-epoch ambiguity resolution for short baselines using combined GPS/Beidou system. 41274045..51 0. Astronomical Institute. GPS Solutions 18. L. Wang et al.73 3.4 0. Corresponding assessment experiments are conducted in post-processing mode and real-time mode with up to 2551 km long baseline. Shi. 2013. Hugentobler.. These results are promising.79 5. Dach. On the other hand.21 2. W. improved DSOP solution with only BDS observation can be expected. 2012. BDS and GPS + BDS (unit: cm). The international GNSS service in a changing landscape of global navigation satellite systems. Although satellite visibility and geometry of BDS are comparable to those of GPS after the enouncement of BDS regional service. 2007. C. GPS Solutions 17. Conclusions Although BDS has been providing regional service since December 27.41 0.86 0. Baseline BJF1–HRBN CHDU–HKTU CHDU–BJF1 CHDU–HRBN BDS GPS BDS + GPS N E U N E U N E U 1.05 0 00 02 04 06 08 10 12 14 16 18 20 22 00 Hour of day Fig. 2013.93 0. Y.70 1.2 0. W.. J.45 3D position error (m) 0. Realization of high-precision relative positioning using Beidou regional navigation satellite system. C.17 0.01 1.80 1. X. The results demonstrate that BDS relative positioning can be achieved with DSOP at centimeter and millimeter accuracy level. C.18 0.M.0 Manual. C.

Hugentobler. Montenbruck. GPS Solutions 17... Anaheim. 2013. He´roux.. 475–486. 2001.. et al. Xu. Xu. Adv. Jin. P. Ge. GPS Theory. W.. C. 2014. Odijk. Geng. GPS Solutions 14.. Zhang. M. 103–119. IGSMGEX: preparing the ground for multi-constellation GNSS science. Wang et al. S. 72851S–72851S-72859.... Niell. Berlin Heidelberg. Springer. Berthias... China Satellite Navigation Conference (CSNC) 2014 Proceedings. Performance assessment of single. 2013a. A. 2009. 2008. J.. J.. T.. Park.. 5005– 5017.. X.F. Geophys. Y. F.. Salazar..E. Steigenberger. Jefferson. Guo. Hernandez-Pajares.. 2013.C. Zhao. H. 102. Zhao. H. Shi. R. 2008.124 M. J. P. U. Wang.. Zhao. Laurichesse. Y... C. W. O. GPS Solutions 18. Steigenberger. X. CA. Li.J. Barcelona. Hu. H. Initial assessment of the COMPASS/Beidou-2 regional navigation satellite system. 3227–3246.). 2010. M. D. J. Jiao. Khachikyan. In: International Conference on Earth Observation Data Processing and Analysis (ICEODPA).. Juan.. Teunissen.. Wu. 2012. M. Orbit and clock analysis of compass GEO and IGSO satellites. J. Hauschild... C.. Geod. C. Teunissen.... 2014. S. Jin. Inside GNSS 9. Springer. S. 2014. 2009. Jiao. GPS Solutions 5. 2014. D. (Eds. Geod. Geod. 197–209. P. 335–350. Heflin. Kouba.. Shi. Global mapping functions for the atmosphere delay at radio wavelengths. H.. N. Tang. Recent development of PANDA software in GNSS data processing. Geophys. H. J... et al. 2013b. P. 335–344. velocity and acceleration determination: algorithms and tools. 293– 299. Hauschild.. Shi.. (Eds.J... H. Berlin Heidelberg. 1997... Li.. M. Salazar. P. Zhang... Sensing snow height and surface temperature variations in Greenland from GPS reflected signals. Wu. J. Geod. M. Steigenberger. 1996. Ge... Sanz. Algorithm and Applications. O.. Precise orbit determination of Beidou satellites with precise positioning. C. Montenbruck. Webb. Effects of physical correlations on longdistance GPS positioning and zenith tropospheric delay estimates. F. 211–222. Najibi.. J.H. III. Precise relative positioning using real tracking data from COMPASS GEO and IGSO satellites. Teunissen. J. J. M. Watkins. Kouba. J.F..-P. O. Zumberge. Ren. Li. D.. pp. 393–403. Precise point positioning for the efficient and robust analysis of GPS data from large networks. Luo.G. Shi. Guo. 42–49. J. Bakker. P. Adv. 55. Precise point positioning using IGS orbit and clock product. Yang. second ed. GPS observations of the ionospheric F2layer behavior during the 20th November 2003 geomagnetic storm over South Korea. In: 2009 International Technical Meeting of the Institute of Navigation. Liu. J.. GNSS multi-carrier fast partial ambiguity resolution strategy tested with real BDS/GPS dualand triple-frequency observations.. Lou. U. Guo. 199–208. 87.. 2013.G. Lu... 251–263. 1079–1086.. 515–525. vol.. P.. slips.. Real time precise gps constellation orbits and clocks estimation using zero-difference integer ambiguity fixing. J. P.. Technical University of Catalonia. O. Res.. M. China Satellite Navigation Conference (CSNC) 2013 Proceedings. Precise GPS-based position. Modeling and performance analysis of GPS/ GLONASS/BDS precise point positioning.. Q. China Earth Sci. 2010.B. Liu. He. 883–892.. J. Deng.. GPS Solutions 17. 46. Geod.. 1623–1633. 2014. Liu. Res. et al. 83. In: Sun. A... Berlin. Shi. J. 2013. 2007. G. 101. J. Nischan. 190–195. Odolinski. Li. GPS Solutions. J. Nakamura..... Single-receiver single-channel multi-frequency GNSS integrity: outliers. and ionospheric disturbances. Space Res. . J. Department of Applied Mathematics IV and Applied Physics... A. C.. W.. 51. Sci. Initial results of precise orbit and clock determination for COMPASS navigation satellite system. Springer. D. J. Q. 1008–1018.M. pp. Triple-frequency carrier ambiguity resolution for Beidou navigation satellite system.... Luo. R. He. J. O. Y. GNSS data management and processing with the GPSTk. Jin. Li. pp. Li. Space Res. P. H. J. A. 88.. pp. The GFZ realtime GNSS precise positioning service system and its adaption for COMPASS. Yang. Z.. / Advances in Space Research 55 (2015) 113–124 J. Montenbruck.. 664–672. 2010. Instantaneous Beidou + GPS RTK positioning with high cut-off elevation angles.. Hugentobler. 2014.. J... C. Zhao. 2013. Wickert. 87. Q. Testing of global pressure/temperature (GPT) model and global mapping function (GMF) in GPS analyses. Xu. 87. Space Res. 53. Adv..). M. D. J.. GPS Solutions 18. Q. J. 12–28.. S... Mercier. Geod.and dual-frequency Beidou/GPS single-epoch kinematic positioning. 82. P.. 161–177.