GPS/GNSS based train position locator for railway signalling

Transcription

1 GPS/GNSS based train position locator for railway signalling A. Filip, L. Bazant, H. Mocek & J. Cach Czech Railways, DDC SZT Laboratory ofintelligent Systems, Pardubice, Czech Republic Abstract Recently, a real-time accuracy of the differential Global Position System (GPS) receivers achieved a sub-metre level in code mode, and centimetre level in the Real Time Kinematics (RTK) mode. Thus GPS receivers became promising for low cost signalling concepts and other safety critical applications in the railway industry. This paper presents a verification methodology and experimental results of the US's GPS and Russia's Global Navigation Satellite System (GLONASS) based train position locator tests at Czech Railways. The verification methodology is derived from the current needs in signalling, and from the parameters of the satellite navigation systems. A key element in the investigation is a switch, on which reliable and continuous position determination of a routing train is most critical. Two-dimensional and onedimensional routing detection models are analysed and experimentally investigated on the switch-point. The trials were performed on Pardubice - Hradec Kralove - Chocen line with the total track length of 100 km. Three mobile platforms were employed for the tests: 1) an electric locomotive operating passenger train for the long term locator verification, 2) a Diesel track motor-car for tests on a laboratory track in Pardubice station, and 3) a remotely controlled ultra-light motor-driven track rover for the detailed locator tests. The data from/to the vehicles were transmitted along the entire trial track though the intelligent 150 MHz TDMA radio modem network with data rate up to 19.2 kb/sec. Finally, the safety-related requirements for future European navigation system GALILEO are specified and discussed.

2 1228 Computers in Railways VII 1 Intelligent trains It is expected that the beginning of the new century brings several new concepts to the railway industry, which increase railway safety standards, operational efficiency and quality of passenger services. These concepts are mainly based on a penetration of the Information and Telecommunication Technology (ITT) into the railway environment. One of these concepts is to build signalling and train control systems preferably free of track-side equipment. The goal of the concept is to transfer a control-decision process from a trackside to a train-born systems, and make so called distributed intelligent train control systems. In other words it means to make trains more intelligent. The concept in final effect will lower investment, operational and maintenance cost, and thus makes railway industry more competitive to the other transport modes. A key element of the concept is a satellite navigation based train-born position locator system, which is able to determine position of the train anytime and anywhere on the track with required accuracy, availability, integrity risk, time-to-alarm, and other important parameters regarding safety applications. Current standalone GPS and GLONASS satellite navigation systems don't meet these strong safety-related parameters. Not even integration GPS/GLONASS with inertial navigation systems (INS) doesn't satisfy these strong parameters. However, a promising candidate for this new railway concept seems European navigation system GALILEO (GNSS-2), to be developed within This paper is the contribution to the definition of the railway safety related requirements for GALILEO system. Commercially available GPS/GLONASS receivers have been employed for the vehicle position determination experiments in order to demonstrate their applicability in signalling. A detailed analysis of the experiments results in the specification of railway needs related to the application of GALILEO system in signalling and train control systems. 2 Satellite navigation in signalling 2.1 Railway needs Numerous ITT technology suppliers offer now to the railway industry relatively cheap standalone "non-differential" GPS train position reporting systems for regional single-track lines in order to "increase railway safety". The suppliers usually support their business proposals with the following arguments. The accuracy of the reporting system several tenths of meters is sufficient in order to stop the train in the case of potential accident on the track between stations. Even if the position of train is temporally unavailable by masking of GPS signal, the system reports the last position of the train before it entered the "dark" area. Thus dispatcher is able to "estimate" the track section on which the train is located. In spite of these "arguments", most of the railway operators still resist and refuse the proposals. Why? They know there are also some stations on the line. And dispatcher must exactly know on which of the parallel tracks in station the train is located.

3 Computers in Railways VII 1229 Currently used signalling techniques. \d procedures guarantee these functions on a certain safety level, even though the signalling equipment is obsolete and overcame. The above "commercially" available GPS based train position reporting systems are not able distinguish with high probability ( %) on which of the parallel tracks the train is located. Further, integrity, availability, time-to-alarm, and other safety-related parameters are completely omitted in these systems, and therefore they cannot be used in safety critical applications. It is evident that the operator of the satellite navigation system must guarantee parameters of the system, which are important for railway safety applications. Up to know such service hasn't been offered to railways. Nevertheless, good quality differential GPS/GLONASS receivers are commercially available now, including inertial sensors. Therefore in these initial R&D phases, the train position locator satellite navigation based can be tested independently as a part of "dispatching" tool over current signalling. This strategy, in contrast to the previous one, is for most of railway operators acceptable. It is desirable if the safety related applications would be developed before GALILEO will be put in operation, i.e. before Detection of train routing at switch-point The most important function of the train-born position locator is its capability to provide very a reliable information on instant position of the routing train at switch-points. Two basic routing detection models are discussed in next sections. Two-dimensional one, which is related to 2D position determination, and onedimensional one, which employ information on heading for the detection. It's evident the information must be provided by a system based on a fail-safe architecture Two-dimensional model Manufactures of GPS equipment usually specify the horizontal accuracy of GPS receiver at 95% of time, it means 2 a (2drms) in Gauss distribution. For example, for GPS receiver with 1 meter accuracy (2 a), one can derive the three following relations: 3 a - accuracy of +/-1.5 m at % of time, 4 a - accuracy of +/-2.0 m at % of time, 5 cr - accuracy of +/- 2.5 m at % of time. Thus the accuracy corresponding to 5 a is taken as a minimum value to get sufficient probability for detection of train routing. Table 1. - Accuracy of GPS receiver - 2 a versus 5 a. Accuracy (2drms)[m] Accuracy (5 a) [m] Table 1 shows a relation between 2 CT and 5 or accuracy's for different receivers. The minimal distance between axis of parallel tracks is about 3.8

4 1930 Computers in Railways VII meters. The usual distance at CD ranging from 4.5 to 5 meters. Therefore, it seems that satellite positioning would satisfy the strong requirements and distinguish that the train is located on one of two parallel tracks with high probability. However, it is necessary to add that the accuracy of 1 meter can be achieved under the following conditions: good satellite visibility (6 or more satellites visible), good satellite constellation above the horizon (PDOP less then 1.5), receiver must operate in differential mode, etc. In case receiver of ordinary and cheap receiver which achieve accuracy of about 15m (2drms) it is evident, that two-dimensional model cannot be applied for routing detection. In order to use not so precise GPS receiver (2drms accuracy of 15 meters) for signaling applications, the one-dimensional model was proposed [1] One dimensional model The one-dimensional model is based on the idea that train runs only on rails (derailment is not considered as an ordinary operation). It means that train runs along beforehand known trajectory (surveyed e.g. with cm accuracy), and can change direction its movement only at exactly known points - switch-points. Since GPS receiver outputs excepting information on position also heading, the heading was proposed as a basic information for detection of train transition from one track to another one. From all routing situations, the most difficult is the detection of routing between two parallel tracks. The minimal length of the transition trajectory between two parallel tracks is about 30 meters. Along this relatively short trajectory, GPS receivers and INS must give the answer if the train transits from one track to other one or not. The below experimental results give you the answer, which of the models is acceptable for signalling. 3 Verification methodology In order to experimentally verify the above presumptions and requirements regarding satellite based train routing and train position determination, it is important to determine the actual position of the test vehicle very precisely and independently on satellite positioning. It is very important since then the GPS based train position locator can be tested on both 1) trial track segments with good satellite visibility, and 2) trial track segments with limited or no satellite visibility. The verification methodology proposed in this paper is based on an interpolation of the precisely surveyed track segments using mathematical analytical formulas. The instant and continuous reference position of the vehicle on the trial track is then calculated using information on travelled distance provided by an odometer. The reference position is calculated in x,y coordinates. Figure 1 illustrates an example of the trial track layout. The trial track is divided into number of segments, which are interpolated by direct lines and circular arcs. The formulas, which interpolate reference trajectory for forward and backward directions, are derived in Appendix. The formulas are derived for the track gradient equals zero. The positive or the negative track gradient must be introduced into the analytical formulas to avoid an error. If the real shapes of the

5 Computers in Railways VII 1231 track are more complicated (such as switch-points), they can be divided into higher number of segments and approximated using polynomials. In the most three parallel tracks Legend: x,,y,...surveyed points on track axis travelled distance measured using odometer R, radius of track arc... centre of track arc Figure 1: Analytical description of the trial track. cases, a polynomial of 3-rd order is sufficient. The accuracy of the interpolated track of+/-2 cm can be achieved. Special markers of known position must be installed on the trial track to switch the calculation of the test vehicle position using one set of formulas to the other one. In order to minimise an error in the calculated reference position caused by an error in the travelled distance measurement, additional track markers can be installed on the track. It is only recommended on track segments with length acceding one hundred meters, m the case of train routing detection on switches, no additional markers are required. The methodology is applicable for all shapes of track including different kinds of switch-points. The advantage of the method consists in the possibility to determine the continuous position of test vehicle on track with sufficient accuracy: less than +/-5 cm in transversal direction and less than +/-20 cm in longitudinal direction. Further, the calculation of vehicle position can be synchronised with the internal clock of GPS (GPS/INS) train position locator. Then the calculated reference and measured positions (in x,y co-ordinates) are provided exactly at the same time. This is very important for detailed verification of the train position locator parameters.

6 1232 Computers in Railways VII 4 Trials In the past, CD performed a number of the differential GPS/GLONASS based position determination trials using different track vehicles. The practical experience resulting from the trials is the following. Long term operational tests on track with the length of several tenths of kilometres are needed for studying of the satellite visibility along the track, verification of the protocols and data formats of the mobile radio network, observation of the operational and the environmental effects. On the other hand, GPS/GLONASS trials on short tracks segments with switch-points are important for R&D in the field of signalling applications. 4.1 Long-term operational tests In order to perform long term operational tests, CD have equipped electric locomotive (3 kv DC) with the differential GPS locator (see Fig. 2). GPS/GLONASS antenna 150MHz radiomodem antenna (a) (b) Figure 2: DGPS train position locator installed on electric locomotive, (a) the antennas on the roof of the locomotive, (b) the locator box. The locator consists of Ashtech's G-12 differential receiver (90 cm accuracy at 95%) and 150 MHz/19,2 kbs radio modem with RF output power of 5 Watts. The train-born equipment also includes extra opto-electronic odometer, which is used for other GNSS/INS trials performed at CD. The locomotive daily operates passenger trains on electrified line Pardubice- Hradec Kralove-Tyniste nad Orlici-Chocen with the total length of the trackage

7 Computers in Railways 171 of about 100 km. The track axis of the line has been surveyed using the differential combined GPS/GLONASS GG-24 receivers. The reference track maps were created using WINPRISM postprocessing software. The track axis has been mostly surveyed with the transversal accuracy of +/- 3 cm. The accuracy of +/- 30 cm has been achieved on some track segments with limited satellite visibility. The rest of track segments with very bad satellite visibility couldn't be surveyed at all. Therefore classical track maps with the accuracy of +/- 20 cm have been used Hradec Kratov6 station Station tower \ _., (height of 40m) i Trebechovice \ Engine N^ * / pod Orebem \ house " \ Tyniste nad Orlici 8km /22 km e krnl Pardubice station / TDMA radio network A Borohradek 19,2 kbs/ mhz / RF output power 5 W / CD Laboratory of / intelligent Systems 1 1 km Station tower (height of 30 m) / Key: # - Locations of radio-modem basestation at railway stations along trial track Chocen Radiomodem outage resistant power supply Figure 3: Radio network for differential GPS/GLONASS train position determination trials. Along this line, the dedicated TDMA radio network has been installed, which consists of 9 base stations, for bi-directional data transmission between the locomotive and the laboratory. The architecture and the basic parameters of the radio network are shown in Fig. 3. The presented radio network is only used for R&D work, since no other railway data radio network is available now. In the future, GSM-R seems promising candidate for this kind of applications on European railway network. G-12 onboard receiver receives RTCM-104 correction signal from Ashtech's GG-24 GPS/GLONASS reference base station to correct systematic errors in position. The train locator transmits NMEA-183 GGA message back to the laboratory. The GGA message includes such information as information on locomotive position, number of received GPS satellites, PDOP, age of corrections and other data. The train-born locator can be remotely configured

8 1234 Computers in Railways VII through the radio network and thus other messages can be transmitted to the laboratory. All the data are recorded in the laboratory and analyzed off-line. 4.2 Detailed tests Since the electric locomotive daily operates passenger trains, it is not allowed to disturb anyhow the engine driver during his work. The locomotive locator hardware and software can be only modified during short periodic inspections and maintenance of the locomotive. Therefore, CD staff have been used for their R&D work, excepting the locomotive, two other vehicles: Diesel track motorcar, and specially built electrically driven track rover (see Fig. 4). The rover is driven by an electric motor with power of 750 Watt and remotely controlled with 35 MHz RC controller. The rover has been used for trials on not so frequently used industrial line with length of 5 km in Pardubice. Figure 4: The rack rover employed for the routing detection trials at switches The trial track has been covered by RTCM-104 correction signal for both code measurement (message type 1, 31) and very precise RTK measurement (message type 18, 19). The trial track consists of segments with very pour reception of satellite signal, mainly track arc in the deep cutting with length of 600 m, and segments with good satellite visibility at switch-point and triple track area. While the "dark" track section has been used for GPS/INS locator trials, the "open" track has been used for investigation of GPS/GLONASS routing detection described in section 5. The "dark" arc track section was surveyed with two differential GPS/GLONASS receivers, which were located on bridges and a steam pipeline bypassing the cutting. The points on axis of track were projected using a plumb line onto the points surveyed with the GPS/GLONASS receivers.

9 Computers in Railways All the trial track segments including the switch points were surveyed with the accuracy of +/- 3 cm and interpolated using mathematical formulas as described in section Equipment The track rover, as shown in Fig. 4, is equipped with two odometers (see Fig. 5). The first one is used for GPS/GLONASS/INS data fusion trials. The second one is used for calculation of the reference position of test vehicle independently on satellite navigation. The differential instant position and heading is measured using the first GG-24 GPS/ GLONASS receiver. The second GPS/GLONASS/GNSS satellites DGPS/RTK reference base station Laboratory site Mobile platform - track rover Figure 5: Experimental set-up. GG-24 receiver operating in precise RTK mode (+/- 2 cm) provides additional positional information on track segments in "open" area only. This information is important for verification of the GPS/GLONASS routing capability. Similarly, the additional information on heading is provided by the KVH fibre-optic gyroscope. The RTCM-104 reference signal is transmitted from the laboratory to both GG-24 receivers through the radio network. The reference data on position and heading are computed using analytical formulas (described in section 3). 4.4 Objectives and procedures The presented experimental results were achieved during the operational tests using train-born differential GPS locator installed on electric locomotive, and using the portable equipment (see Fig. 5) installed on a track motor-car, and a special track rover.

10 Computers in Railways VII Main attention was paid to the experimental verification of GG-24 routing detection capability on the switch-points. Further, the observation of GPS and GLONASS satellite visibility, accuracy in position, and other parameters under different conditions on tracks were investigated with respect to the future applications in signalling. Two routing models have been experimentally investigated: two-dimensional and one-dimensional. In the case of two-dimensional model, the track rover was running over the switch-point, and its position measured by GG-24 receiver was recorded. The measured trajectory in code mode was compared with the computed reference trajectory. Similarly, the instant heading in dependence of the travelled distance was recorded and compared with the reference heading calculated from the reference trajectory. Heading data were also provided by KVH laser gyro with data rate of 10 Hertz to verify its functionality for the routing detection. Since the GG-24 receiver provides position in WGS-84 co-ordinate system, the data had to be transformed to x,y projection. 5 Experimental results In this paper, there are presented the results illustrating the routing detection for the following configurations: a) mixed mode (GPS+GLONASS) differentially corrected with the age of RTCM-104 corrections of 1 second, b) GPS differentially corrected with the age of RTCM-104 corrections of 50 seconds, and c) the instant heading provided by thefibre-opticgyroscope. Excepting this, the advantage of the combined GPS/GLONASS receiver is demonstrated in the case of the track surveying in the deep track cutting. Finally, the results concerning a number of received navigational satellites on track at hilly areas on single-track lines and stations, and on double-track corridor lines are also presented. 5.1 Routing performance The axis of the tracks entering/rising to/from the switch-point were surveyed using DGPS and postprocessing with accuracy of +/- 2 cm and approximated using cubic polynomials to get the reference trajectory. The reference heading in dependence of the travelled distance was derived from the reference trajectory. However, there are very small differences between two boundary points of two adjacent cubic parabolas, which ranging from 1 to 2 cm. Thus the reference trajectory is not a continuous curve and some small step changes appear in the reference heading. However, the changes are negligible from the viewpoint of the verification methodology. The time delay, which was intentionally introduced into the generation of the RTCM-104 correction signal by the reference GG-24 base station, simulates a degradation in the accuracy of the mobile receiver.

11 Computers in Railways VII Two dimensional routing model - position measurement 1237 The positional data was selected from NMEA-183 GGA messages, which also include average number of received satellites (SV), PDOP, the age of corrections (t), etc. The parameters are described in each graph reference B-HD meas. track 1 A--A meas. track , x[m] Figure 6: Measured position of the track rover at switch-point, (GPS+GLONASS, SV=10.5, PDOP=U, 1=1 s). Figure 6 shows measured position of the track rover at the switch-point according to the configuration a). The accuracy in position less then 1 meter was achieved. As results from Tab. 1, it is possible to distinguish that the vehicle is located on one of the parallel tracks with probability greater than % of time. The measured trajectory according to configuration b) is in Fig. 7. The accuracy of the measured position was degraded to about 10 meters. This accuracy is not sufficient the for train routing detection within the twodimensional model. The age of the correction of 60 seconds degrades accuracy to 15 meters level, and it is not applicable for the train routing at all. From other experiments, which are not graphically presented in this paper is clear, that the age of corrections up to 15 s didn't introduce an observable degradation of the GG-24 receiver accuracy in GPS mode. The age of corrections of 30 s degraded the accuracy to 2 meters. The age of RTCM-104 corrections of 5 seconds is recommended in order to save radio channel capacity.

12 1238 Computers in Railways VII reference B EI meas. track 1 ^--6 meas. track x[m] Figure 7: Measured position of the track rover at switch-point, (GPS, SV=7.63, PDOP=1.4, r=50 s) One dimensional routing model - heading measurement The GG-24 outputs the heading in NMEA-183 VTG message. Figures 8 and Figure 8: Measured heading of the track rover at switch-point, (GPS+GLONASS, SV=10.5, PDOP=1.1, t=l s).

13 Computers in Railways VII 1239 illustrate the measured heading by the GG-24 receiver according to the configurations a) and b), respectively. Figure 10 illustrates the measured Figure 9: Measured heading of the track rover at switch-point, (GPS mode, SV=7.63, PDOP=1.4, t=50 s). Figure 10: Measured heading of the track rover at switch-point using KVH gyroscope. heading using the KVH gyroscope. In case a) the measured heading is applicable for the routing detection. If the age of corrections of 50 seconds was 100

14 1240 Computers in Railways Vll introduced, which corresponds to about 10 meters in accuracy, the measured heading was degraded too much it cannot be applicable for routing the detection (see Fig. 9). As results from other tests, the maximal age of corrections of 30 seconds is acceptable for the routing detection. KVH gyro provides very good reproducibility in the heading measurement on the switch, as shown in Fig Satellite visibility along track GPS receiver on a single-track line inflatarea and passing through a forest can usually receive from 3 to 5 satellites. Sometimes, GPS receiver cannot calculate _1525 E, >% 1520 B -a A -A track GPS+GLONASS GPS r r x[m] Figure 11: The difference between track surveying using GPS and GPS+GLONASS (GPS SV-4.5, GPS+GLONAS SV-7.5). position (only 2 or 3 satellites visible), or the differentially corrected position is degraded from 1 meter to several tenths of meters. If the satellite signal is blocked by trees without leafs, the GPS receiver can receive signal from 2 to 3 more satellites. The situation is nearly the same at hilly areas with deep track cuttings and high hillsides along the track. However, there are more sites on track where GPS receiver cannot calculate position. GLONASS usually add from 1 to 3 visible satellites. At stations, where yard area is wider and sky more open than a forest path one single-track line, 6 or more GPS satellites are visible. The situation is practically the same at stations atflator hilly areas. GLONASS again add from 1 to 4. It is the important fact for the detection of train routing. On double track corridor lines the satellite visibility is nearly the same as at stations. Although, the current GLONASS constellation consists of 11 operational satellites only, its effect is still positive in railway applications. Mainly for track surveying in deep track cuttings, as illustrates Fig. 11.

15 Computers in Railways VII i Conclusion The experimental results presented in this paper show that signalling and other railway safety critical applications require GPS receiver with accuracy of 1 meter. This receiver provides both the positional and the heading information, which can be used for detection of train routing at switch-points. Excepting this, other railway applications require 1 meter accuracy for train position determination, such as shunting operations and precise train stopping control along a platform. A stand alone one-dimensional routing detection model [1] is rejected in this paper. GPS receiver accuracy of 10 meters is not able to provide reliable information on heading for signalling applications. On the other hand it is recommended to employ a combination of one and two-dimensional models for GPS and INS integration. However, GPS receiver with accuracy of 1 meter must be used. It is clear that current GPS and GLONASS don't meet safety requirements. In spite of this satellite systems are sufficient for current development of railway safety related applications, which will be based on GALILEO system in the future. Which basic railway safety related requirements should GALILEO satisfy? First, it is clear that current coverage by the signal from GPS/GLONASS satellites is not sufficient to meet railway safety requirements. The GNSS-1 will be put in operation this year but doesn't solve the problems with the coverage. Although thefinalconstellation of new GALILEO system hasn't been defined yet. It should bring at least 24 new navigational satellites. Under these presumptions it is evident that GPS receiver on train will be able to receive at least 6 satellites on single-track line. The same receiver will be able to receive at least 12 satellites on yards. It is sufficient number to detect train routing. Together with the additional inertial sensors, obviously. Although GNSS-1 with EGNOS overlay satellite system is able to provide differential corrections through the satellite link, this function doesn't seem critical in railway safety-related applications. The RTCM-104 correction signal can be transmitted from a ground GPS reference station to the trains through train radio data network, which is necessary for other signalling functions. The data quantity in the RTCM-104 message is 380 Bytes in worst case and it can be compressed to about half. If the compressed message will be sent to each train every 5 seconds, as it proposed in section 5.1.1, there is a capacity enough to perform other signalling functions. The efficiency of the distribution of the correction signal via satellite link (GNSS-1) will be experimentally investigated within DG-XIII's APOLO project at RENFE and CD this year. CD has already specified the other safety-related parameters for GALILEO system. The parameters are the following: availability greater than 99.99%, integrity risk at least 3.0"^ [I/ hour], continuity risk 4.0~^/30 second, time-to alarm less then 1 second. Future GALILEO providers should guarantee the safety related-parameters to railway end-users within a special service.

16 I ~>42 Computers in Railways Vll Appendix The following formulas approximate the continuous reference trajectory along the track axis (see Fig. 1). The formulas are derived for forward and backward directions of movement. Linear track section 1-2 Forward direction: where Backward direction: *2,i = *2 + ^2,1 cos(#j; ^,i Circular track section 2-3 Forward direction : *2,3 = ^ + *i cos(^ ^2,3 =Cy +R^ sin(^2,3 ) where Backward direction : 2)= where References [1] Differential GPS: An aid to positive train control. Federal Railroad Administration, Report to the Committees on Appropriations, June 1995.