Metro Traffic Regulation and Quality of Passenger Service

Abstract

The Paris metro network operated by RATP is well-known all over the world for its complexity and high level of efficiency. Since most of its equipment dates back from the 1970’s, it is undergoing an important modernisation plan, among which the renewal of all the equipment involved in traffic control, both on the lines and in terminals, within the Operational Control Centres (PCC). This new PCC system has now been implemented on lines 13 and 3, and although the operators of those lines are generally pleased with the technical features of the new PCC, they remain somewhat doubtful regarding its efficiency on automatic traffic regulation. The objective of this project was to detect if indeed those mixed feelings were founded by studying the specifications of the new system as well as its impact on operations through an analysis of the operational data. Although it was not a primary objective of the PCC project, it is hoped that this new system can help improve traffic regulation performance, and through this the overall quality of passenger service which is very much related to how regular the traffic of a metro line is. Public transport in Ile-de-France is governed by the public authority STIF. Every four years, a new contract is signed between STIF and RATP regarding operations in the metro network. This contract defines the total number of kilometres that need to be produced every day of the year on each line, but it also sets quality indicators among which two are directly related to traffic regularity: passenger waiting time and production during peak hours, the latter having been introduced in 2008. Production and quality of service are rewarded or penalized financially. RATP therefore has strong incentives to both produce the contractual number of train*kilometres, but also to ensure a regular traffic on their line. In order to operate the lines according to those contracts, theoretical timetables are built according to a number of constraints such as headways to be implemented, running times, dwell times, buffer times at terminals. Although those timetables are never communicated to the passengers who are only given the maximum waiting time depending on the time of day, they are the basis of operations in the metro lines. In reality, it is almost impossible for the metro lines to respect those theoretical timetables because of all the incidents, but also because it is very difficult to predict the time needed to be spent at stations. With more passengers, the trains needs to stay longer in stations, thus creating additional delay on the whole line. The later a train is, and thus the more space there is between it and the previous train, the more passengers will be waiting for it, and the more time it will have to stay in the stations. The traffic on a metro line as busy as the ones in the Paris metro network are therefore highly unstable and a single incident can completely disrupt the traffic on the whole line for a very long time. This is especially true of lines where trains are running very close to each other (down to 95s at peak hours on line 13). Because of this, it is very important to implement the appropriate regulation strategies. At RATP, the main objective of the regulation actions is to maintain a regular interval between trains at departure and on the lines while still trying to recover from delay in terminals. Before the new PCC system was implemented, traffic on the line was overseen by the traffic controllers who were all located in the Bourdon building and used button-control panels, while departures from terminals was overseen by the departure chief in the terminals. In the new system, not only have the equipment been renewed, operations on the line have also been re-organised and the regulation strategies are now the responsibility of the computerized system. Traffic regulation on the line is still the responsibility of the traffic controller, but he is now located at the decentralised PCC, closer to the lines. With him are also the two terminal managers who now oversee all train movements in their terminals from the PCC. As for the calculation of departure times, whereas it was before the role of the departure chief, it is now the computer that calculates them according to the general decision made by the terminal manager who can choose between a no-regulation mode, a constant headway mode and a recovery mode. The latter is based on the former strategy that consisted in distributing the delay among trains in order to ensure regular departures in the line when there was a disruptive train causing delay at departure, and headway reduction in order to recover from the delay when the disruption has disappeared. When it is not possible anymore to recover from delay with those recalculation methods, namely because the accumulated delay is too important, the last resort of the terminal manager is to cancel some departures, thus leading to the so-called “lost laps for schedule compensation”. After studying the specifications of the algorithm for the calculation of departure times, it was concluded that they were in line with RATP’s requirements. However, those requirements date back from almost ten years ago, and the needs of the line operators have evolved ever since then, especially with the introduction of the “production during peak hours” criteria, which might already explain the mixed feelings of the operator regarding the system’s regulation functions. Although the operators are not completely satisfied with the system’s performance, it needs to be stressed out that they are overall quite happy with the new equipment. Furthermore, the new organisation ensures a much better communication between the various actors of traffic regulation which is much appreciated. However, traffic controller do feel a loss at being now separated from the other lines, and the terminal managers at being separated from their terminals for they do not have direct communication with the drivers anymore. As for the actual impacts of the system on operations, it was possible to conduct a performance evaluation using the operational data provided by the IGT and OSIRIS software, specifically developed for RATP. Among the various methods that were considered for this data analysis, the most efficient one was to assess the number of lost laps for schedule compensation as opposed to the number of incidents, or to the total delay on the line per day. Indeed, on days with the same amount of delay, more cancelled departures mean that the system is not able to recover from delay alone, and is therefore less efficient. Various periods before and after the implementation of the new system were thus compared. The first conclusion of this study was overall that the situation on the lines had not particularly been improved with the implementation of the new system, neither had it worsened. Although this conclusion is already quite satisfactory, as it was never an objective to improve the operational results of the lines, it still needs to be taken into context. Indeed, the operational environment of the lines nowadays are very different from what they were even 5 years ago for the passenger flows and the number of scheduled trains have increased, and other modernisation project have been or are being implemented at this time. It is very likely that the old system would not have been capable of handling those changes, and would not have had as good operational results as the ones observed with the new system. Unfortunately, this is only a guess that can never truly be verified. It was also observed on both lines 13 and 3 than the implementation of the new system was always followed by very difficult periods for the operators and that problems regarding parameterization had occurred quite often, especially on line 3. A simple model of the calculation of the algorithm on line 3 was thus built in order to try and study the impact of parameter changes. The outcome of this study was not only that a more precise calibration of parameters was necessary to improve the efficiency of the system, but also that the main calculation of the algorithm could very well be modified in order to better fit the evolving needs of the lines. A first recommendation for the improvement of the system is therefore to conduct a more extended study on this option. Furthermore, the specific characteristics of the lines and the evolving needs and structures should be better taken into account. Finally, another improvement could come from a better training and support of the operator who, as it is, does not have a good enough level of understanding of the system, which leads to misinterpretations and misuse on his part.