Location and positioning in the railway (part 2)

GNSS

There are now multiple GNSS (Global Navigation Satellite Systems) constellations available for our use. The original one Navstar (commonly known as GPS – Global Positioning System) was developed by the US in the late ‘70s. This is now supplemented by Glonass (Russia), Galileo (Europe), Beidou (China), QZSS (Japan), IRNSS (India). In the UK for instance, we typically get good coverage from GPS, Glonass and more recently as newer receivers support it, Galileo is becoming operational.

GNSS works by having line of sight to a constellation of satellites orbiting in space, to achieve the best accuracy with GNSS, our receiver needs to be able to see a good spread of the satellites (we need some above, to the left, right, ahead and behind us ideally), this can be quantified mathematically by a value known as dilution of precision. The most used form of this is called PDOP (Position dilution of precision), this is a numerical measure of the satellite geometry (the spread of the satellites as the receiver sees them) - the smaller this value the better.

The images below give an idea of what is a good and bad PDOP, imagine the receiver is at the centre of each set of circles. The top image shows a spread of satellites in every direction - this is a good PDOP (low value), the second has satellites ahead and behind the receiver but none to the side - this will give a high PDOP. With a low PDOP there is more information that enables a receiver to calculate position with a lower ambiguity than with a high PDOP.

PDOP is a big topic that we won’t cover here - but the important factor in the railway is that it is very hard to achieve a low PDOP all the time, embankments, tunnels, dense urban areas, OLE and other parts of infrastructure block the signal - meaning that the PDOP usually changes frequently and is often high (more like the second image). With a high PDOP, it is unlikely that we will have a high accuracy position.

PDOP should not be used as the sole source of GNSS quality – it is a guide. Other factors can be at play, for instance, PDOP is a measure of the geometry of the satellites that the receiver is receiving signals from. This doesn’t mean that all of these satellites are in the direct line of sight to the receiver. Multipath is of concern in the railway, multipath is when a signal is reflected and/or refracted from objects near the receiver. The trains we use, generally operate in high multipath areas (in effect longitudinal canyons a lot of the time). High quality (particularly multi frequency) receivers and antennas can significantly reduce multipath errors.

The spec sheet of a typical survey grade receiver will state its capabilities, something like an accuracy of 1cm + 1ppm. What does this really mean? This means that in these ideal situations where we have a very low PDOP (great view of the sky in all directions) can we expect to get a position with an accuracy of 1cm? Not quite, this also includes another figure, 1ppm - this is an additional error based on the distance of the receiver to a reference station. Again, in here, we don’t want to go into the technical details of this but to get these accuracies we also need a correction service such as RTK (Real time kinematic) or PPK (Post process kinematic), these provide correction data using ground-based reference stations that are often miles away. 1ppm means one part per million and equates to an additional error which is a function of the distance of our receiver to that (or those) reference stations(s).

There are lots of other considerations in how RTK (Real Time Kinematic) or PPK (Post Process Kinematic) work and how their accuracies are calculated and maintained, for instance, both require tracking of satellites for an amount of time (they are not instantaneous), a change in the status of satellites being tracked may change the quality of the RTK/PPK solution - and therefore the accuracy. PPK is usually more accurate than RTK as there is access to more information, for example, once an RTK fix is obtained, the accuracy obtained applies from that point in increasing time until a change in the tracking status (of satellites and/or the correction), whereas PPK can “re-wind” the fixed correction back in time because it is a post process operation. But of course, neither of these give great results in challenging GNSS areas (urban canyons, tunnels etc) where the signal may be lost all together, this is why these methods are often supplemented with an INS (Inertial Navigation System), these devices are able to bridge the gap between areas of GNSS coverage - giving continuous position where no satellite signals are being received. Depending on the grade of device, the drift (the rate the error increases) away from the truth could be a few centimetres for hundreds of metres - of course, this depends on what the truth is and whether the device is operated in real time or post process.

To decide on which positioning system should be used for IM data, it is first essential to decide what latency is required in the data. For instance, if it’s acceptable to have a latency of (say) 24-48 hours after a shift, then a PPK INS solution is most likely to be suitable (given an upper limit as to what is an acceptable positional error of course). However, if results are needed in near real time (this could be within X minutes of measurement or even quickly after the end of the shift), then it is likely that a PPK INS solution will have too high a latency and as such, we will have to look for another solution, most railways now use some form of GNSS/INS either with a correction service (such as RTK or SBAS (Satellite Based Augmentation System) such as Terrastar) - or without, relying on uncorrected positions alone. Both options can and will produce erroneous positions in certain scenarios and without quality control and correction, this could cause issues for IM.