Research Brief Obtaining data to assess the dependability of GNSS information and accuracy of odometry T510 - December 2009 Background In recent years a number of research projects (such as APOLO, LOCO, GADEROS, RUNE, ECORAIL and LOCOPROL) have investigated various aspects of the application of global navigation satellite system (GNSS) technology, such as GPS, to the railway domain. These projects have confirmed that, when suitably aided with other sensors, it is functionally possible to use GNSS technology to determine train position and speed. Whilst these projects have demonstrated that it is, indeed, feasible to use GNSS technology for position and speed determination on the railways, the dependability of the information from these systems has not have been addressed. To meaningfully support the introduction of GNSS technology to the GB railway, the dependability of the information provided must be assessed alongside the accuracy of train odometry systems. This research provides the reassurance that GNSS information and odometry accuracy is a reliable means of determining speed and positioning of trains on the GB rail network. Aims The aims of this research were to enable the theoretical understanding of the dependability of GNSS derived position and speed to be verified consider the augmentation and use of current odometry systems for the odometry function for ATP on low density and rural lines. Method The research designed a GNSS data collection system for installation on a train in normal revenue service. This equipment collected GPS and odometry data over a 12 month period. The collected data was processed and analysed and the dependability of the position and velocity obtained was assessed. A sub-set of this equipment was fitted to an additional train, with different wheelslip/slide protection and odometry systems. This train was used to collect odometry data during a three month period in the leaf fall season. This data was analysed to assess the performance of the odometry systems during their most RSSB R&D Programme Block 2 Angel Square 1 Torrens Street London EC1V 1NY research@ssb.co.uk www.rssb.co.uk/research/ rail_industry_research_programme.asp 1 Obtaining data to assess the dependability of GNSS information and accuracy of odometry T510 - December 2009 challenging periods of operation using the same techniques as used for the main data collection. Two trains (A Class 323 operating in the Birmingham area with London Midland and a Class 170 operating in East Anglia with National Express East Anglia) were equipped with identical sets of equipment containing a number of different GPS sensors, an inertial measuring unit (IMU) and GSM modem together with benign interfaces to the train's speed measurement system and the cab occupied signal. Once the driver's key had been inserted in any cab of the train formation the equipment started logging information. Data collection on the Class 323 started in November 2007 and in March 2008 on the Class 170. The analysis of the data was performed between the start date and December 2008 for both trains, although data collection continued for longer than this. The data was correlated against data obtained from the Ordnance Survey's national network of GPS stations that continually record raw GPS data. The data from the dual frequency, survey quality GPS receiver on the train was processed to give accurate locations of the train and also used for an independent assessment of the quality and accuracy of the GPS system. The GPS data collected on the two trains was then analysed in terms of the ability of the GPS system to provide a positioning solution, including analysis of the distribution of the duration of any outages. The number of satellites tracked by the receiver and also used by the receiver for the generation of the position solution was also obtained and this was used to determine the ability of the GPS system to provide not just a solution, but a solution with integrity. The raw GPS measurement data that was recorded was also used to generate information regarding the variation of the received strength of the satellite signal and the effects that this could have on the positioning solution. The investigation of the odometry provided by the two different speedometer systems was performed using the stand alone GPS velocity as the source. To analyse the distance function a simple velocity integrator was used and the speed was integrated between pairs of stations to obtain the distance between the start and stop points. The analysis of this data provided information on the suitability of the distance derived from the speedometer data as a source for the odometry function for train positioning on rural routes. 2 RSSB Deliverables The research has delivered: A final report, aimed at the specialist technical reader, who needs to understand the project and its methodology, and how satellite navigation performs in the railway environment. A recommended specification for GPS equipment on trains, aimed at the more general technical/managerial reader, who may need to make decisions when specifying satellite navigation equipment. A database of raw GPS and odometry information; collected during this project, and which could be of value in other research in the rail industry. It is available for other investigations through application to RSSB. Findings The findings are inevitably complex in nature and the full report should be consulted to understand the findings in their entirety. In summary, this research provides the reassurance that GNSS information and odometry accuracy is a reliable means of determining speed and positioning of trains on the GB rail network. GPS dependability The availability analysis showed that in the urban area around Birmingham four satellites could be received for approximately 85% of the time, reducing to 75% of the time if six satellites were required to provide autonomous satellite based techniques for the integrity of the GPS system. In rural East Anglia, four and six satellites are available for over 95% of the time. The distribution of the length of the outages observed was that over 85% of the outages in the Birmingham area lasted five seconds or less and over 90% lasted ten seconds or less. The effect of the environment around Birmingham New Street is seen in that approximately 4% of the outages last for three minutes or longer. The average duration of an outage computed for the Class 323 was found to be just under ten seconds. The Class 170 showed outage statistics that were significantly better than the Class 323, because of the flat nature of East Anglia with very few large obstructions. The number of outages that were five seconds or less was over 98%, and around 99% for 10 seconds or less. The average outage duration was found to be just over two seconds in length. For the Class 170, out of a total of almost 56,000 recorded outages only 14 lasted longer than 50 seconds RSSB 3 Obtaining data to assess the dependability of GNSS information and accuracy of odometry T510 - December 2009 and only 64 lasted longer than 20 seconds. It is believed that these occurred on the rare occasions that the equipped unit worked in to London Liverpool Street. The analysis of the stand alone GPS positional accuracy, for the Birmingham area only, yielded statistics showing that the performance of GPS in the urban environment was quite good, with maximum positional errors in 95% of the "route bins" being less than 105m based on data from the entire year. From the analysis of the SNR information and the positional accuracy information, there appears to be very little seasonal effect on the accuracy of the position solution generated. Skyplots clearly showed the obscuration caused by deep, steep sided cuttings and reinforced the theoretical models. Odometry The odometry experimentation suggests that, after GPS calibration, either speedometer system could be used to provide accurate distance measurement, however these results are based on data sampled at 1Hz over a limited time period. It is possible that there are occasional undetected excessive errors present in the system because of the low sampling rate or the relatively short data collection period, which gives some uncertainty as to the intended ATP application. From the experimentation performed, the following results were obtained for the Class 323 and Class 170 units. Differences in distance between the GPS and speedometer derived values for the class 323 were within ±10m for over 90% of the time for the vast majority of the inter station distances computed. There were however areas where the performance was worse, but these tended to be in areas where there were fewer journeys or the GPS coverage was poor. If a difference of ±50m between the two derived distances is considered the worst percentage seen to lie within this range for the Class 323 was approximately 92%. A difference of 50m on the average journey length for the Class 323 of 1.74 miles is an error of about 1.8%. The Class 170 odometry experimentation yielded similar statistics to the Class 323, although the range of differences was larger with around 90% of the values lying within ±50m for most of the journeys. It is believed that this is due to the longer journeys for the Class 170 in comparison to the Class 323 together with a higher operating speed and potentially a more sinuous route between the station stops. A difference of 50m on the average 4 RSSB journey length for the Class 170 of 11.40 miles is an error of about 0.3%. Next Steps V/TC&C SIC has reviewed the recommendations and agreed that the next steps resulting from the T510 project are that: Based on the results obtained, Network Rail will derive a statistical model describing the different GPS error distributions seen and assess these derived models in terms of the different applications for which the use of GPS has been considered. This aspect is encompassed in the LiveTrain project. The research results will be used as an input for the future review of GE/GN8578 'Guidance on the Use of Satellite Navigation' to parameterize performance in different railway environments. Contact For more information please contact: Head of Engineering Research R&D Programme RSSB research@rssb.co.uk RSSB 5