Comparing Manual Driving and Driverless Operation
Section 2.1.2 described the different Grades of Automation (GoA) defined by UITP, with manual driving of trains in GoA 0 and 1, and automatic train operation (ATO) in GoA 2, 3 and 4. The benefits of introducing ATO over manual driving are more consistent performance: closer working to line speeds and avoidance of ‘over-cautious’ station and signal approaches , which in turn can support greater network capacity. The aim of the simulations carried out for this study was therefore to compare the capacity of the Tyne and Wear Metro under manual driving (GoA 0 and 1) with ATO (GoA 2, 3 and 4—which therefore covers driverless trains).
The Tyne and Wear Metrocars have camshaft control of their electric traction equipment, and as such the maximum acceleration is always demanded. The ATP system on the Metrocars is intermittent, protecting against signal overruns but not providing continuous speed supervision. As such, manual driving can maintain the average cruising speed at close to the line speed limit. The key difference between manual driving and ATO for the Metrocars is therefore during braking, where ATO can provide a higher average deceleration than less consistent (and more cautious) manual driving .
The average in service deceleration achieved with manual driving during braking of the Metrocars has been measured at 0.5 m/s2 by previous work, well below the full service braking figure of around 1.3 m/s2 . For ATO, the London Underground Central Line can be used as a point of comparison. This uses a target deceleration of 0.75 m/s2 when running above ground and 1.15 m/s2 in underground sections. The full service maximum here is also around 1.3 m/s2, and this provides an illustration of the differences in deceleration levels for manual driving and ATO referred to above. However, when the Central line ATO system was first introduced, there were significant issues with low wheel/rail adhesion levels on above-ground sections, with wheel slide under braking resulting in unacceptable levels of wheel flats and station overruns. Low adhesion conditions on the line are most commonly reported during autumn, typically due a combination of light rain and fallen leaves on the rails. When these conditions are present, the ATO target deceleration is reduced to 0.55 m/s2, in addition to other mitigation measures such as railhead treatment .
The use of ATO makes it more practical to implement moving block signalling [22, 32], and the two technologies are frequently implemented together on driverless metro systems around the world. Therefore, the replacement of the existing fixed block lineside signalling by a moving block signalling system was also considered for the simulation work described in this paper.
This study builds on a previous experience with rail simulation work at Newcastle University. The Metro has previously been modelled using discrete event-based simulation software packages, such as Simul8  and Arena . For this study, continuous physics-based simulation software was required to predict the effects of changing deceleration levels on inter-station journey times and investigate the resulting changes to the interaction between trains, and hence determine the influence on capacity. OpenTrack multi-train simulation software (v1.7.5) was used for the modelling and simulation work. A model of the Metro system had been built and validated against measured speed profile and energy consumption data in a previous project. For this study, the capacity of the central corridor between South Gosforth and Pelaw was analysed. The boundaries of the study were therefore set to be Regent Centre, Longbenton, Hebburn and Fellgate, so that the constraints of the flat junction at South Gosforth and the single line section to Hebburn were included.
Landex  discussed a number of different ways to define capacity, and subsequently detailed the standard UIC 406 method of measuring it . This method has been widely adopted for studies of capacity in metro systems . The theoretical maximum capacity of a line is first calculated by compressing the timetable—i.e. reducing the headway (separation) between trains to the minimum possible. A timetable based on this maximum theoretical capacity would have no resilience against delays however, and as such would not be practical for day-to-day operations. Real timetables therefore include additional time margins that increase the headway between trains, which gives the practical capacity of the line for a given level of timetable stability.
The current (2016) Metro timetable has a maximum of 20 trains per hour (tph) through the central core at peak times, and this was implemented in the OpenTrack model as the baseline for practical capacity. The simulation was based on the assumption of an ‘all-out’ driving style to maximise capacity, with maximum acceleration, cruising at line speed and then braking, with no coasting. The current train lengths were assumed to remain the same. The maximum theoretical capacity was obtained by reducing the headway between trains in the model to the minimum possible before OpenTrack identified signalling conflicts starting to occur. Figure 2 illustrates this timetable compression process for trains running between Jesmond and Central Station under the existing fixed block signalling. In this example, reducing the headway any further would create a signalling conflict at Jesmond. The minimum headway for the overall timetable was used to evaluate the ratio between practical and theoretical capacity, which worked out as 76% for the current timetable.
The rolling stock deceleration specified in the OpenTrack model was then modified to examine the effects of ATO in accordance with Sect. 4.1, and the new maximum theoretical capacity obtained by the same compression method as before. This new maximum capacity value was then multiplied by the 76% ratio for capacity utilisation derived above, and rounded down to the nearest integer value to provide an estimate of practical capacity (in tph). The OpenTrack model was also modified to replace the existing lineside signalling with a moving block signalling system. This capacity estimation process was repeated for all of the combinations of manual driving/ATO and lineside/moving block signalling. Low adhesion conditions were also tested for the ATO case, with both lineside and moving block signalling, again by modifying the deceleration value in accordance with Sect. 4.1.
An additional benefit of ATO identified in Sect. 3.2.1 was better recovery from delays, some of which is derived from the parallel use of moving block signalling . This means that it could be possible to reduce the additional time margin between trains, while still maintaining a given level of timetable stability. This would effectively increase the capacity utilisation above the 76% ratio derived for the current timetable with manually driven trains. A sensitivity test was therefore carried out using ratios of 80, 85 and 90% for the various ATO cases to examine the possible increases in the number of trains in the timetable.
As well as changes to signalling and control, Nexus are also considering new rolling stock, with a higher maximum speed of 100 km/h . The current Metrocars have a maximum speed of 80 km/h, and the camshaft control results in a sawtooth profile for the tractive effort at low speeds, providing an average acceleration of around 1.15 m/s2. The OpenTrack model was therefore modified to also test this case. The new rolling stock characteristics were based on the London Underground Central Line rolling stock, with a maximum speed of 100 km/h and a constant initial acceleration of 1.3 m/s2, although the maximum power rating of the traction equipment was assumed to be the same as the existing Metrocars. Where the current line speed is 80 km/h, it was assumed that this could be raised to 100 km/h in the model. Much of the central core of the Metro is restricted to lower speeds however, typically due to line curvature, and these restrictions were left in place. All six combinations of manual/ATO/low adhesion ATO and lineside/moving block signalling were tested with the revised train performance and line speeds.
One of the assumptions in the timetable modelling was a nominal dwell time of 30 s at each station. The dwell time can potentially have a significant effect on the capacity of a metro system [24, 50]. As such, a second sensitivity test was carried out for the capacity in the baseline case of manual driving and lineside signalling, using dwell times of 15, 30, 45 and 60 s.
There are three remaining assumptions in the modelling that are not yet detailed. The safety distance for the moving block signalling system was assumed to be 150 m, identical to the safety overlaps in the existing fixed block signalling. Train mass is assumed constant, based on a typical passenger load. The Metrocars have load weighing in traction and braking, so the effect on journey times (and hence capacity) of variation in mass is minimal. Finally, the power rating of the overhead line equipment and substations is assumed to be sufficient to support additional trains. The power supply system would require an additional investigation to determine the changes required with a revised timetable, but this is out of scope of this study, as the focus is on the potential capacity increases made possible by ATO.
Results and Discussion
The results for practical capacity through the central core section of the Tyne and Wear Metro are given in Table 3, in terms of trains per hour (tph). The use of either ATO or moving block signalling in isolation allows an increase from the currently timetabled 20 to 22 tph, but implementing the two technologies together allows an increase to 30 tph. As such, both should be implemented together to obtain the maximum capacity benefits from these technologies.
The figure of 30 tph is reasonable by comparison with what has been achieved on other urban rail systems across the world . Although resignalling is one of the prerequisites to achieving these benefits, moving block is not an absolute requirement, as it is possible for in-cab fixed block signalling with shorter blocks to approach the capacity provided by moving block.
However, Table 3 also indicates that low adhesion conditions are sufficient to negate nearly all of the capacity benefits of ATO/moving block signalling, with the necessary mitigation measures identified in Sect. 4.1 reducing the capacity from 30 to 23 tph. On the London Underground Central Line, the core section is almost entirely underground, and the majority of the adhesion issues are encountered at the outer ends of the line where service density is typically lower. On the Tyne and Wear Metro, a significant proportion of the central core section is above ground, including a number of tree-lined cuttings, which are one of the most problematic areas for low rail adhesion. Measures to mitigate low adhesion may include vegetation management, fitting sanders to rolling stock, rail head treatment trains or more radical options such as linear motor technology [51, 52]. Conventional (adhesion-worked) railways make up a minority of GoA 4 metro systems in Table 1, and it is only recently that they have become more common; the risk associated with low adhesion conditions are likely to be partly responsible for this trend.
Increased Capacity Utilisation
The results of the sensitivity study for increasing the capacity utilisation ratio are given in Table 4, for ATO with existing signalling and ATO with moving block, for both normal and low adhesion conditions. The size of the increase for increasing the capacity utilisation ratio is significant.
Further detailed investigation of a specific signalling/control system and timetable would be required to determine whether an increase in the capacity utilisation ratio would be possible while still retaining an acceptable level of timetable stability however.
Higher Performance Rolling Stock
The higher performance rolling stock had little effect on these results, typically only changing the headway by around 2–3 s. This is likely due to the close station spacing and number of speed restrictions in the central core of the Metro, which means that there is little opportunity to take advantage of an increased top speed. The advantages of this increase in top speed would be seen in journey times on the outer parts of the network, where there is greater distance between stations and fewer speed restrictions. The increase in initial acceleration also made little difference, as it is for a relatively short duration while the train is accelerating; more time is spent in the constant power region of the tractive effort curve. An increase in the maximum power of the trains would make more difference to capacity, but the equipment cost and energy consumption would increase. The relative increase of 1.15 to 1.3 m/s2 in acceleration is also rather less than the relative increase of 0.5 to 0.75 m/s2 in braking.
Sensitivity to Dwell Time
The results of the sensitivity study on dwell time for the existing manually driven trains and lineside signalling are given in Table 5. Reductions in dwell time below the nominal 30 s assumed for the modelling appear to provide a small increase in capacity, but increases in dwell time can result in large reductions in capacity.
For comparison, testing a 60 s dwell time in the moving block signalling/ATO case reduced the capacity from 30 to 22 tph. Poor design of the platform–train interface and poor management of passenger flows within the train and within the station therefore have the potential to negate the capacity benefits of investments in vehicles, signalling and power supply if dwell times become the critical factor.
The simulation results also provide journey times and energy consumption of the trains between South Gosforth and Pelaw (and vice versa), and Table 6 illustrates the relative differences in journey time and energy consumption between manual driving and ATO, for both the existing and the higher performance rolling stock. These results are for the existing lineside signalling.
As noted in Sect. 4.2, the simulation results assume all-out running. The trade-off between journey times, energy consumption and overall capacity illustrated in Table 6 can be altered by changing the driving style, for example, by introducing coasting [34, 35]. Likewise, increasing the maximum power rating of the higher performance rolling stock above that of the existing Metrocars will also alter this trade-off. The choices in rolling stock and signalling design, the timetable and the operation of the network are ultimately a balance between many competing factors that aim to optimise the benefits of the system in relation to the costs.
Accuracy/Reliability of Results
The multi-train simulation software measures journey times and headways to a resolution of one second, and the results of the validation exercise carried out previously suggest an overall accuracy of around 3 s for station-to-station journey times. Comparison against the differences in headway for the driving/signalling options considered implies that this accuracy is sufficient for the main conclusions about the capacity benefits of driverless trains to be valid, and also suggests that the differences in capacity from the higher performance rolling stock tested are not significant.
Further development of this work would be to move to more detailed investigations of specific timetables for the entire Metro network, including studies of the likely dwell time at each station, and Monte Carlo simulation of pseudorandom delays to estimate actual timetable stability . This would provide a sound basis for the development of new commercial timetables to take advantage of the potential capacity increases offered by driverless trains.