Making every second count
SWITZERLAND: SBB is focusing on making better use of reserves in train scheduling to add greater robustness to its integrated national timetable. Real-time management of train services to maintain connections and more detailed passenger information provision are at the heart of the ‘Co-Production’ concept.
Dr Felix Laube, Head of Processes & Methods, Infrastructure & Traffic Management Dept, Swiss Federal Railways
Marco Lüthi, Swiss Federal Institute of Technology
If rail transport is to meet the demand for sustainable mobility, it must provide its passengers with fast, cheap, safe and reliable services. At the heart of Swiss Federal Railways’ Bahn 2000 concept is an integrated symmetrical periodic timetable offering excellent connections at major hubs throughout the network (RG 12.01 p838).
However, the success of Bahn 2000 has brought the network to saturation point. Predicted growth in demand will be impossible to handle with existing planning and operation principles, and our current operating policy relies increasingly on breaking connections to stabilise the network.
To improve operational performance while satisfying customer expectations, SBB has started to re-engineer its operational process through a ‘Co-Production’ concept. This approach allows intelligent reserve allocation by combining Japanese-style process control with real-time scheduling.
Public transport, and especially rail, has played an important role in Switzerland for a long time. In 2003, market share for public transport was 21% for passenger and 34% for freight. SBB operates a 3 000 km standard-gauge network, virtually all of which is electrified. A further 2 000 km is operated by around 50 private railway companies on standard and narrow-gauge lines, also almost entirely electrified. The network is finely meshed with connections at numerous nodes.
The railway is a true mixed traffic network: based on UIC statistics from 2006, SBB runs an average of 93·8 trains/day on each route, making it the most heavily-utilised rail network in the world. At the same time, the quality and punctuality is remarkable when compared to most other European countries. In 2006, 96·2% of all SBB passenger trains arrived with a delay of less than 5 min.
Bahn 2000 was developed around the aim of connecting the entire country with an integrated clockface timetable. Main stations are served at regular intervals, and rolling stock and infrastructure management is co-ordinated to maximise the number of connections at each interchange. The integrated timetable provides for optimal timed transfers and results in high accessibility and generally shorter journey times.
The bulk of the Bahn 2000 timetable was implemented in December 2004, and most routes are now operated on a 30 min frequency throughout the day. The service has been extremely successful in attracting more passengers to rail. From December 2004 to December 2007, overall demand for SBB passenger services increased by 38%, and the trend is still accelerating. Demand is expected to increase further as additional elements of Bahn 2000 are completed.
However, this success has brought its own problems, as the network is now nearing its capacity limit. Consequently, the service risks becoming more unstable and connections are missed as a result of very small delays.
As part of its research into optimising operations, SBB recorded one passenger’s experience of travelling from Burgdorf to Basel via Olten one evening in June 2007: ‘the scheduled connection time in Olten was 3 min. All was going well until we got close to Olten, when the train started to get slower and slower until it finally stopped and waited. This is the connecting passenger’s nightmare. This delay lasted more than 3 min and there was no announcement in the train as to what was happening. Eventually a freight train passed by slowly and then my train started again. It must have made up about 1 min. I positioned myself at the train door, running as fast as possible to the platform for the Basel train. As I got to the top of the ramp, 15 sec after the scheduled departure time, the train was already moving. So, just because I was 15 sec late, I had to wait half an hour … in a station where all restaurants, shops and toilets were closed’1.
Our research suggested that SBB had to focus its development on two areas. First, a more customer-oriented production methodology was needed, focusing in particular on assuring connections and information management. Secondly, the existing infrastructure had to be used more efficiently to meet surging demand and to optimise costs.
In a system with an integrated clockface timetable, the importance of reliability over pure punctuality is evident. Delaying a train a little to make a connection is often better than insisting on a perfectly punctual departure, as long as delays can be made up during the journey and are of no material significance for later services. Equally, it is totally unacceptable in such an extensive and interconnected network to have no further information available beyond arrival and departure times at particular stations.
SBB thus took on the task of improving control of the existing reserves within the operational process, and everyone involved in the process (dispatchers, infrastructure managers, drivers and guards) had to revise their approach to planning and operations.
The provision of reserves in the timetable influences the performance of the network. On one side, such reserves reduce the efficiency of infrastructure utilisation and unnecessarily extend travel times; on the other side they make the system more stable and robust to cope with delays or unforeseen events. The key therefore is to allocate the allowances intelligently.
Reserves can be split into three categories:
Tolerances ensure that those responsible for individual process outcomes are able to pursue their tasks successfully, for example train drivers must be able to follow their designated timetable paths without affecting other trains. The tolerance range is a hard constraint for path generation. As long as the train remains within its tolerance bandwidth, the path, and by consequence the schedule, is guaranteed to be conflict-free.
Dwell time and time supplements to allow recovery from delays. A two-stage algorithm allows the targeted placement of dwell time and recovery allowances. On a global level, the optimal allocation of time windows can be determined using a flexible periodic event scheduling model. On a local level, the time windows are exploited to find the most efficient resource allocation in each capacity-constrained zone. The remaining time can then be optimally distributed along the line in compensation zones using a resource time conflict graph2.
Unallocated capacity Spare paths are intentionally kept free of train movements and ready for allocation if necessary.
To ensure an efficient and effective operation, the three core service production processes (on-time departure, on-time driving and short-term changes to the operating plan) have to be controlled precisely. SBB’s Co-Production concept combines dynamic real-time rescheduling and accurate production and is designed as a superposition of two feedback control loops (Fig 1).
The main idea is that every train always has an up-to-date and conflict-free schedule detailing time, speed and route information. These schedules have an accuracy measured in seconds and are continuously available for all the actors involved in a train’s operation. Fast algorithms are needed to generate revised schedules whenever a train exceeds a tolerance threshold or other scheduling conditions change (for example through changing train dynamics or infrastructure failure).
By referring back to an abstraction of the timetable called ‘service intention’, the algorithms ensure that connections are not broken unnecessarily and customer-oriented decisions are taken. An intelligent time-distance discretisation known as PULS was introduced by Roos3, which simplifies the possible choice of options and generates revised timings for capacity bottlenecks within the shortest time. Fig 2 shows a time-distance graph for the area around Luzern.
In order to make most effective use of the improved rescheduling process, train operators must be able to follow new timetables more precisely than they could before. This requires a train control tool that can help drivers to follow very precise dynamically-changeable schedules. One possible approach is to use a driver-machine interface, which works as a controller to help the drivers follow a new trajectory in the event of disruption. However, the trains would still remain under the full control of the driver to ensure safe and timely running.
The departure process must also satisfy strict time restrictions. To reduce inaccuracy and large statistical variations during the departure period, enhancements are needed to both technology and process management. One possible solution is the parallelisation of processes to reduce the duration of the departure, and more detailed information systems which can give passengers more exact departure times.
The area around Luzern is being used as a pilot project area to research the Co-Production method. Various sub-processes were simulated and tested during regular daily operations before all the elements of the concept were applied to a series of trains arriving and departing simultaneously. The first tests were undertaken at the end of 2007 and in early 2008 to gain feedback from everyone involved and to gauge the performance of the technical equipment.
The results so far have been promising. Drivers have welcomed the driver assistance tool as an aid to on-time running, and the deviation in departures from Luzern was limited to 15 sec.
In future, railways stand to gain by viewing operations not as a sequential, event-driven process, but rather as a time-driven one. Improved management of reserves causes a leap in process predictability. Passengers get more reliable outcomes and unpopular dispatching decisions, in particular missed connections, are minimised. If and when these do occur, reliable alternatives can be offered to each passenger, tailored to their individual needs.
This will help to improve customer satisfaction and strengthen rail’s competitive position. In addition, the improved control will allow operational resources to be allocated in a more targeted way. Thus, existing infrastructure can be utilised more efficiently and effectively. Additional benefits include reduced energy consumption and improved staff morale. All this helps to reduce costs and allows for further demand to be met within existing infrastructure.
However, it is important to understand that the change in process management and methodology cannot be enacted simply by ‘buying tools’. A change in attitudes towards continual improvement of processes is required at all levels of the organisation.
- Fig 1. The Co-Production control loop concept.
- Fig 2. A PULS diagram showing real-time changes to the pre-defined fixed time-distance pattern for the Luzern area.
- Fig 3. The Co-Production concept allows real-time adjustment of train pathing via a driver interface in the cab. This allows dispatchers to alter arrival and departure times at key hubs, as shown in this graphical representation of an inter-city and regional service making an adjusted connection at Luzern.
- Kenworthy J. Making Connections, SBB AG Bern, Internal Report. 2007
- Caimi G, Chudak F, Fuchsberger M and Laumanns M. Design of a new railway scheduling model for dense services, Proceedings of 2nd IAROR Seminar, Hannover. 2007
- Roos S. Bewertung von Knotenmanagement-Methoden für Eisenbahnen, Master Thesis, ETH Zurich. 2006.