Location and positioning in the railway (part 6)
Additional possibilities with RailLoc
Change detection
As the infrastructure changes, the features used in the RailLoc map do too. This could involve a feature gradually moving (as bits of ballast shift and settle) or a feature no longer being present. Both can lead to additional benefits:
Tamping - the change after tamping allows RailLoc to provide information about the extent that was tamped (exactly) and where it was NOT tamped
Changes due to renewals - where a section of track has been renewed, this can be picked up straight away using RailLoc and fed back into the Network Model as well as being used to invalidate previous geometry recordings (for example - as the geometry has changed, previous defects may no longer be valid)
Monitoring of ballast - for gradual changes of ballast, either through reduction in features or perhaps movement of features, RailLoc may be able to provide insight where these phenomena are unwanted (e.g., voiding)
Input into signalling systems
As mentioned before, systems such as DAS/ATO/ETCS often consume data from balises installed on the physical infrastructure. This is a potentially costly exercise requiring time on site (boots on ballast) as well as planning, surveying and maintenance. These are safety critical systems relying on the availability of all the sensor inputs, additional redundancy can be gained through consuming data from RailLoc - this could be continuous or on a low frequency basis (acting as a virtual balise for instance). Given sufficient frequency of operation, RailLoc could be used to increase the spacing of balises (i.e., reduce the number required and therefore reduce the installation and maintenance costs). This would be done through fixed updates (i.e., virtual balise) but also through higher quality continuous data (supplementing or replacing tacho and GNSS) into these systems.
Summary & conclusions
In this series we have discussed the typical approaches to railway positioning and particularly those used for positioning Infrastructure Monitoring data.
The devices and systems we have discussed have stood us in good stead, but there is more we can do. By embracing newer technologies, we can replace manual processes, retain data and be more efficient. Ever increasing demand on the railway means that the need for better data and automation has never been higher, increasing capacity and frequency of services means that we need to keep boots off ballast where possible. The safety implications of workers being on a live railway mean that we should only send them to site where it is absolutely essential.
We have discussed, inertial navigation systems (combinations of GNSS and IMU) which are now commonplace, these can give excellent results, but not everywhere in the railway. The railway does not lend itself to perfect GNSS positioning because of the large number of occlusions of the signals (embankments, OLE, tunnels, station canopies etc.), this means that we either accept a degraded or incomplete dataset or we look to supplement or replace this approach.
GNSS based systems can produce higher quality data through additional post processing, but this causes delays to the solution that may not be acceptable for IM data. This approach still doesn’t give us the answer everywhere. We have also discussed further sources of error such as the underlying network model (both topology and geometry) which can lead to attributional errors.
Additional sources of positional correction are available, but these are likely to either require installation to physical infrastructure (as well as maintenance) or offer a lower (spatial and temporal) resolution. In fact, RailLoc can be used so supplement or even replace these.
The advent of computer vision has unlocked more innovate approaches to positioning, such as RailLoc. Rather than supplement GNSS based systems with external information, RailLoc provides the base truth and uses GNSS/INS for infill. Unlike the other systems discussed, RailLoc works everywhere on the railway and requires no installation to physical infrastructure. RailLoc can not only give excellent relative positional accuracy but can provide high absolute accuracy through initial survey. RailLoc is also able to provide certainty that a defect (or full path) was captured on a specific track with categorical assurance.
Away from positioning data on vehicles, we have also discussed the issue of track worker navigation to defects and suspects captured by these systems. Standard handheld GNSS devices can work well, but struggle in challenging GNSS areas and their errors can be large (or worse misunderstood). Building on the RailLoc functionality, Fault Navigator allows navigation of a maintenance worker to the precise location of a fault, significantly reducing time on site and repeat visits.
And finally, when considering accuracy, correct attribution of line is often overlooked in favour of a +/- value. It is perfectly possible for a position to be within the prescribed accuracy and still be totally incorrect due to attribution. Consider an ultrasonics suspect, a positional error of 10m may still lead to the suspect being found - but only if the worker searches on the correct line.
