Knowing a train’s location is a vital piece of information in the control of train movements – one that has existed almost since railways were first born.
In the earliest of days, time interval working was used, where trains were dispatched at set times in the hope that the second train would not catch up with the first, but, after a few nasty accidents, something else was needed. Thus, a form of train location system was devised.
Advances in technology over the years have led to a number of systems being developed. These can be listed as:
- Absolute Block Working – a train’s location is known to be between two adjacent signalboxes, often several miles apart.
- Track Circuits – the rail-wheel device that will detect the presence of a train by the wheels providing a short circuit across the rails. Track circuits can vary in distance and can be hundreds of metres in length, so the train location is only known between the track circuit ends.
- Axle Counters – a more-reliable replacement for track circuits. However, they also can often count in and out over a long section of track.
- Induction Loops – two wires laid out between the rails, with periodic crossover positions, to give reference locations. They constantly transmit information to and from the train, usually associated with Automatic Train Operation (ATO).
- Satellite Tracking – technology derived from military and automotive systems, where a train aerial constantly receives the geographical location and displays this either to the driver or is transmitted onwards to a control office. Does not work in tunnels or other covered areas.
- Camera Images -a forward-facing camera ‘compares’ the actual image of a train’s position against images held in a reference data base. The resultant position can be transmitted to a control office.
- Acoustic Sensing – A train’s vibration pattern as it progresses its journey is picked up by lineside sensing equipment, usually a fibre optic cable. The resultant change in optical patterns will constantly detect a train’s presence and speed.
All of these have strengths and weaknesses. The original requirement of interfacing with the signalling equipment to allow the clearance of signals or the setting of routes is clearly vital in terms of safety, so such devices are invariably SIL4 rated (safety integrity level 4).
There is also a need to ensure that a train is complete (that a coupling has not broken) and the safety-based location devices achieve this. However, these devices are less able to provide the precise position of a train as it journeys forward.
With the ever-increasing demand both to optimise performance and to make expert judgements on re-timing trains when things go wrong, knowing the exact position and speed of a train at any point in time becomes essential, especially when penalty payments are involved. This requirement has resulted in the adoption of modern technology that feeds performance systems rather than safety applications.
There is also the cost factor. Infrastructure providers and train operators want value for money and, if modern technology systems prove to be significantly cheaper than the traditional detection devices, then they are likely to be adopted.
To try and bring all these factors into a single perspective, the Institution of Railway Signal Engineers organised a webinar in late February, during which suppliers could present their products and vision. It proved to be a fascinating session, if only to demonstrate the difficult choices that have to be made.
Track circuits and axle counters
Track circuits have been around for decades and have progressed from simple DC battery-fed circuits with insulated rail joints, through AC power-fed circuits of 50Hz, 125Hz and 331/2Hz frequencies to provide immunisation from traction systems, to higher-frequency devices that enable tuned circuits to be established and eliminate the need for rail joints. Rated as SIL4, they are high up in the list of safety requirements and will be around for many years yet.
However, the variable resistance of track ballast in wet and dry conditions, as well as the vulnerability of the wired connections to damage by on-track machines, make for reliability problems. Richard Hinson from London Underground stated that track circuits were the biggest cause of all failures in the signalling equipment portfolio. This situation, coupled with problems in obtaining spare relays for the older-generation equipment, suggests that track circuits are no longer the favoured system for train location requirements.
Axle counters are the logical alternative. Equally troublesome when first introduced several years ago, design and configuration improvements now make them the system of choice when safety considerations dominate. Early problems with miscounts and the lengthy time for resets to restore normal operation have largely been overcome by the building in of intelligence features that distinguish between a train and an unwanted disturbance. Modern designs are clamped to the rails rather than bolted to them, which would require the rails to be drilled, in an improvement that meets with approval from track engineers.
Manfred Sommergruber from Frauscher described the mechanical strengthening that has been built-in to the company’s latest product (SENSiS) to combat climate change, dirt ingress, flooding, rail hammering effect and deliberate vandalism. Connecting axle counters to the signalling system has been made easier with the adoption of digitisation and the replacement of relay interfaces with a serial data-stream. Not only does this make the device more reliable, but it enables more information to be provided, such as wheel diameter and temperature.
Improved diagnostics and the opportunity to use a radio connection are there for the taking, but, if landlines are preferred, connecting all axle counters within an area onto a data ring allows for continuous operation should a cable break occur.
As indicated earlier, both track circuits and axle counters have the disadvantage of only knowing a train’s location between two specific points.
Realisation that pulses of light within a fibre optic cable could detect local vibration and thuds occurred several years ago. Detecting rock falls was an initial application, but, since then, the technology has been developed to provide a means of detecting trains. An injected light source on to a dark fibre will see a marginal change in the refractive index where any disturbance takes place. The time taken for this ‘back scatter’ to get back to the source enables the distance to be calculated.
The processing by a tracking algorithm has advanced to enable more intelligence to be obtained from the reflected pulses, according to Kevin Tribble from Sensonic, and the latest systems are capable of measuring both train location and speed with interfaces to traffic management systems (TMS) and customer information systems (CIS). Installations exist worldwide, including on Network Rail and London Underground.
Several factors need to be understood for an acoustic system to be deployed:
- Fibre location (always assuming a fibre cable already exists and has spare fibres within it) and its installation method – buried, in troughing, laid on the surface;
- Calibration as to knowing the fibre to track distances and whether fibre spools exist in joints;
- Classification as to train mass, speed and size limits, plus performance variables across the tracks.
Ongoing development continues to improve interpretation of the fibre disturbance, resulting in a higher dynamic range that can detect wheel flats, detection of track conditions including broken rails and, most importantly, which track a train is on. The approach has changed from being quantitative to qualitative, and it is foreseen that acoustic sensing will be able to augment ETCS positioning information. The system clearly has much promise and may be able to fulfil both safety and precise location requirements.
With many trains now equipped with a forward-facing camera for security purposes, can this camera be used to identify a train’s location? For some time, Richard Shenton from RDS International has been developing the Valise system (Virtual Balise), where the real-time picture is compared to a stored picture, thus producing a location position. Being entirely train-based and with the camera already installed, this offers a low-cost solution to the challenge but there are potential drawbacks to be addressed.
To have a full and continuous CCTV picture all of the time would require massive amounts of data to be processed. Instead, the stored picture is reduced to a ‘fingerprint’, containing just the essential information needed for the location algorithm. The reduction is around 1000 times, allowing the whole rail network to be contained in a few gigabytes of storage. The small ‘fingerprints’ allow the live picture to be matched to a stored image in real time on a low-cost computer.
In addition, the fingerprinting process provides the robustness to match locations in changing environmental conditions. Weather conditions, including snow, are claimed not to be a problem as sufficient similarity exists between real and stored images. Of course, track remodelling would need the stored image to be updated, so an element of re-work will always be necessary.
Trial results for identifying the correct track from a single image indicate the following performance:
- Normal daylight including rain – 99%
- Night-time usage – 92.5%
- Snow in normal daylight – 75%
Results from a number of image matches are used to achieve the required level of confidence. For ‘along track accuracy’ the position is within 50cm for 68 per cent of the time and two metres for 99.7 per cent of the time.
Confidence and usage would need to be gained incrementally, starting initially with non-safety situations working up to the possibility of SIL2 applications, such as door closure and speed supervision. Compared with GPS, the system has the advantage of knowing which track a train is on.
Trials are currently underway at a location in the UK and a fuller article on the system may appear in due course. For use of the positioning information outside the train, any such production system would need the means to transmit the location data to where it is needed.
With SatNav systems regarded as a normal part of road vehicle equipment, how suitable are satellite systems for train location purposes? Vincent Passau from Alstom gave details of the EU-backed 2020 STARS project (Satellite Technology for Advanced Railway Signalling).
Whilst its prime interest is supplying high-integrity signalling systems, Alstom was looking to use additional technology to overcome some shortfalls in odometry, as used for ERTMS distance measurements. Choices are wheel-based sensors, radar (sensitive to weather conditions), GPS/GNSS (subject to signal availability and multi-path reception), optical rail readers (installation constraints) and accelerometers.
Of these, a satellite-based solution is the most promising, but it needs to have higher accuracy to give Positive Train Detection (PTD). The outcome would be fewer balises plus more information for ATO stopping distances and on-board passenger information updates.
Hence STARS, with objectives to assure predictable performance, interoperability and alignment with the European Shift2Rail initiative. Assessing GNSS (Global Navigation Satellite System) accuracy in field measurements indicates there is a gap between the requirements and results, so it is likely that an enhanced odometry solution will be required as well.
Gyrometers and accelerometers would be used to cover tunnels or other locations where a satellite signal is lost. Precise inertial navigation will be needed to cover short-term changes. Large-scale trials are planned in Norway, with the overall objective of contributing to a SIL4 rated computation and data merging.
It is a given that any ATO system must know the exact location of all trains in the system and have a guarantee of train integrity. Raymond Sturton from Thales gave a brief history of the Seltrac system development, from its initial use of track loops for positioning information to the current deployment of radio using RFID (Radio Frequency Identification) tags placed in the track together with radio antennae. Axle counters continue to be deployed for secondary detection purpose as well as giving assurance on train integrity. Both track loops and radio tags give reliable positioning but hinder track maintenance.
In the search for a train-centric location system, a future NGPS (Next Generation Positioning System) is being developed using ultra-wideband radio (UWBR) that will be positioned at platforms, junctions and other significant rail features and will dispense with the track tags. This, together with radar and LiDAR devices, will achieve accurate location information. The system has no under-carriage installations and no track-based equipment. Trials are underway on the Flushing lines of New York City Transport. UK applications are planned for an Advisory System for Signallers (ASSIGN) on the Barnstaple and Okehampton branch lines in Devon, and as an interface to TPWS Mk4 on the Hertford Loop test track to give continuous over-speed monitoring.
Comparison with road transport
We are all aware of the research into autonomous vehicles and connected transport, which require accurate location and speed data. Raphael Grech from CAV Spirent made the point “if it moves, it must position”. Driverless vehicles will need highly accurate positioning equipment which cannot be achieved with just a single sensor. So, a combination of GNSS, radar, LiDAR, cameras, localised assets (lane positioning) and cabin sensing will all be needed, as will connectivity between all of them. Synchronising position with other vehicles is essential.
Three factors need to be fulfilled:
- Local Positioning – where is the vehicle in relation to local topography?
- Relative Positioning – where is the vehicle in relation to other vehicles and people?
- Global Positioning – where is the vehicle location?
For the latter, GNSS is the only system available, but it gets taken for granted which can lead to wrong decisions being made. Interference, risk of spoofing, segment errors, multi-path connections, atmospheric conditions and cyber-attacks are all being investigated by the military. KPIs are integrity, continuity, robustness, accuracy and availability. Test methods vary, but they must cover everything. Much use is made of simulation but ‘live sky’ testing must happen at some point.
Receiver design is important – if it cannot see a satellite, then detection will not take place and an inaccuracy of one to two metres is unacceptable. Receivers should ideally receive signals from at least five satellites, which need to have different position angles. Constant monitoring as to how the system is working is necessary.
The complexities of logging the position of autonomous road vehicles to the accuracy required make the train location challenge look easy perhaps, but a lot more money is available for research and development?
Some questions and thoughts
Whilst the ideas and analysis of train location systems proved fascinating, from a customer’s perspective, could it be somewhat bewildering? Maybe a potential purchaser needs to consider what the system actually needs to do?
The traditional SIL 4 systems of track circuits and axle counters give information sufficient to set routes and clear signals, but they are of less value when the precise location of a train is required. SIL2 systems, such as acoustic sensing and satellite tracking, give a precise geographic location but may not be able to determine the actual track that a train is on, nor that the train is complete. Camera imaging has the advantage of low cost but an integrity level that would be insufficient for SIL4 applications.
Then there is the vexed question of standardisation versus innovation. If a particular technology was selected, would it need to be adopted on a large geographic scale to ensure interoperability?
A wholly train-borne solution makes this problem somewhat easier. Of all the technologies considered, the fibre-based acoustic sensing seems to offer the greatest potential, as it can do both positioning and train integrity with the ability to count wheels and bogies. With more-intelligent algorithms, it should also be possible to detect which particular track carries a train.
All of this begs the question as to whether a ‘Track Map’ could be defined but, even if it was possible, it would need to cater for the ‘changing face of the railway’, such as weekend renewal work. It seems likely that a combination of systems will be necessary to fulfil both safety and commercial requirements, much as is happening in the road industry.
The next few years should prove interesting.