(without frames    contens    library)

Introduction to the Traffic Control Concept of the City of Zurich

H. Stadelmann, Dr. Th. Riedel and M. Vollenweider

 

1. Overview

The city of Zurich has been implementing a computer aided traffic control strategy since the seventies. The deployment of computers for improving traffic control and quality of public transportation still requires active leadership in the development of new products and components since, unlike in other industries, few innovative, matured products are being offered in the market place.

This article describes the currently implemented traffic control system of the city of Zurich that has been developed in close co-operation with the local industry and the Swiss Federal Institute of Technology. The concept is leading edge in an international comparison. The article first describes the political, financial and technical factors leading to the implementation of the traffic control system. The second part of the article explains and describes the technical implementation and the next steps for further enhancement of the system. The presented solution shows how total operating costs and investments can be reduced and, simultaneously, both the effectiveness of the existing system as well as the flexibility for the fulfilment of the political traffic strategy can be improved in an increasingly more complex and less flexible political and financial environment.

2. Reasons for the development of Zurich’s traffic control system

Flexible fulfilment of the political mission

In a time of a quickly changing political and social environment the flexible fulfilment of the political mission is becoming increasingly important, above all in urban traffic management. Three main factors shape this trend:

Zurich’s traffic control system almost optimally fulfils the above-mentioned basic conditions. Large software flexibility and intelligent algorithms running on extremely cost-effective and standardised hardware infrastructure allow for a simple adaptation to changing conditions and traffic control policies.

Increasing utilisation of existing resources without large investments

Due to the bad financial situation of most cities and shortage of space and due to increasing difficulties in urban road construction (legal proceedings, political pressure), existing traffic systems can only be optimised in existing structures.

Zurich’s traffic control system has been designed under this perspective as a part of the "Urban Road Management": The detailed optimisation of the individual intersection and the entire system allows the optimisation of the overall efficiency without expensive further construction (e.g. over- and underpasses, subways). Of course, the traffic control system can not prevent local traffic breakdowns completely. However, it increases the critical traffic density, so that a breakdown happens not at all or later and congested areas can be cleared more quickly.

Increasing cost efficiency

The total costs of the traffic control system are composed of investments and maintenance costs of the traffic infrastructure, and of investments and operating costs of traffic control departments.

Zurich’s traffic control system not only prevents massive investments in expensive traffic infrastructure, but it provides a high degree of standardisation and automation as well as significant savings in the operating costs of traffic control departments. New control algorithms of individual intersections (e.g. for temporary construction sites) can be drafted and simulated using real-time data on computers in the traffic control headquarters in a fraction of the formerly required time, and can be down-loaded through the network into the corresponding traffic control computer of the modified intersection.

Improving the traffic quality for different traffic categories

An essential advantage of Zurich’s traffic control system is the ability to improve the traffic quality for the different traffic categories. On the one hand, the increased overall efficiency shortens average waiting times of all traffic categories at intersections (e.g. by using adaptive, automated control algorithms) and thus provides shorter travelling times, on the other hand, means of public transportation can be prioritised, which reduces travelling and waiting times and improves punctuality.

3. Advantages of Zurich’s traffic control system

Flexible fulfilment of the political mission

Zurich’s traffic control system provides the technical basis for a flexible fulfilment of the political mission. The system can improve the overall traffic quality and can distinguish actively and adaptively certain categories of traffic. For example, public means of transportation can be prioritised, especially in peak periods, which leads to better observance of schedules and shorter travelling times. This in turn increases the attractiveness of public means of transportation and reduces private traffic. Additionally, lower operating costs can be achieved for public transportation by being able to decrease the number of vehicles travelling due to faster circulation.

Financial Benefits

Lower operating costs

Zurich’s traffic control system is extremely cost effective (less than Sfr. 20'000 per intersection and year). This corresponds to about two employee-weeks including material required. The costs mentioned include software modifications caused by temporary construction sites or changing flows of traffic as well as hardware maintenance and replacement costs.

Reasons for low operating costs are:

Protection of investments

Investments in new generation computers which allow using a large part of the currently installed hardware, are modest. Old components can be exchanged as needed. Additionally, the system is open i.e. Zurich is not locked into a single supplier relationship for software programming.

Modularity

The traffic control system has been designed to modularity (traffic lights, controller, communication line, networks) such that a continuous migration to the new system is possible.

The adaptation of an existing traffic light control system does not require a complete rehabilitation of all existing subsystems. The components can be replaced gradually. The new systems can interoperate thoroughly with the existing intersection controllers, and only the traffic control computer needs to be replaced in a first step. Light-signals, intersection controller, communication lines, traffic control computer and their network are modular components.

Scaleability

A traffic control system can be expanded gradually and is fully scaleable. The deployment of a new traffic control computer is economical from around ten intersections upward. Each traffic control computer can be expanded to control up to 90 intersections. In this way, it is possible to introduce a modern controller generation with relatively modest investments in a step-by-step approach.

The existing personnel of the traffic control department can handle the migration process to the new system since the process is gradual.

Remote maintainability

The costs for software maintenance are very limited, since the maintenance can be done in the headquarters and no on-site work is required (remote serviceability).

Maintenance

The new development environment facilitates also the maintenance of the existing control algorithms, and the time for modifications can be minimised. The time required to design a completely new control algorithm could be reduced greatly, e.g. down to two weeks from up to two months.

Centralisation

The concept of the central traffic control computer minimises costs of peripheral components. This is especially cost effective in systems with large numbers of intersections.

Low-cost communication

The communication lines between the networked computers and between the computer and the intersections are no costly proprietary developments, but composed of low-cost off-the-shelve products. Existing wiring/cabling to and from intersections does not have to be replaced.

Technical benefits in planning and organisation

The technical benefits of the system are closely related to its financial benefits. They rather describe the technical and process improvements than the economic savings.

Open platform

The new traffic control system has been designed as an open platform to allow further developments. Obviously, the further development of controller software is only limited by processor speed of the computer board in the traffic computer and, of course, by scientific progress. The modular structure of the traffic control computer itself enables further development to allow adaptation to future technical developments.

Flexibility

A fast and flexible adjustment and further development of the traffic control system must be based on software adaptations and as little as possible on hardware replacements. The control algorithm can be quickly adjusted to changes of traffic patterns, not requiring expensive and lengthy structural measures at the intersection. This enables easy implementation of new control strategies.

Development environment

The design of a new control algorithm for a particular intersection is done interactively on the computer in a development environment that provides a whole library of control designs, from simple cyclical control algorithm up to new control strategies which have been developed during the last years at the Swiss Federal Institute of Technology in Zurich. The control algorithm can be developed anywhere in the network, even on PCs, and can be down-loaded to the specific intersection.

Testing

A control algorithm can be simulated and tested in detail with the help of a simulator with real or hypothetical data from the particular intersection before down-loading.

The simulator can visualise traffic flows and traffic light control signals on- or off-line and evaluate and correct the control algorithm during operation.

Evaluation of the control efficiency

Simulation and visualisation of control algorithms in operation can be evaluated along various parameters, e.g. average waiting time or number of vehicles in waiting queues. The visualisation can be done on- or off-line on any computer. Data extending over a long period of time allow conclusions on the traffic behaviour or on the efficiency of the control algorithms, and interactive simulation and testing can lead to suggestions for improvements of these control algorithms.

Communication

Each intersection can arbitrarily exchange data with any other intersection. This happens locally inside the computer, if the intersections are controlled on adjacent computer boards (node controllers), or the data can be communicated through the hierarchical network between the traffic control computers. Intersections never exchange information directly thereby keeping networking costs and time delays low and allowing data gathering for evaluation.

The communication allows co-ordination of larger numbers of intersections to systems. It can also be used to track public transportation vehicles (trams and busses) over a longer distance. Especially where tram vehicles of different lines use the same tracks, individual tracking is possible, if the lines divide at an intersection, e.g. the two-way car traffic does only have to be stopped when a tram of line A has to turn left, but not when a tram of line B continues going parallel to the car traffic.

The control algorithms can also communicate through the network with any other application, e.g. programs for traffic supervision, controller evaluation or measurement of traffic flows.

Redundancy and safety

The system is redundant, for each traffic computer is composed of two computing units. If one fails to operate the other one takes over. If only an individual computer board fails, each other free computer board can take over after a short transitional phase during which the traffic lights will be blinking amber.

4. Technical solution

Principles

The technical solution of the traffic control system has been based on the following central ideas:

Hierarchy of the traffic control system

The approach to the intersection

When our car approaches an intersection, odds are that the traffic light is on red, so that some time remains to explain the hierarchy of the traffic control system.

Many intersections are still today controlled by fixed-time algorithms which can be exchanged several times during the day. The car has probably driven over a detector during the approach. The detector sets off a counting signal announcing the arrival to the intersection or the departure from the last intersection. If the vehicle is a tram or a bus the control algorithm extends, shortens or changes the phases of the traffic light at the intersection to let the vehicle pass with high priority.

Detectors detect metal moving over them. They produce signals which are processed to detect as many erroneous signals as possible. However, sometimes detectors tend to oscillate wrongly announcing several cars instead of only one. The traffic management system has the capability of identifying and correcting such errors.

After having received the signal of the car’s arrival, the traffic light transmits visually the decision of its control algorithm by either remaining on red or switching to green.

The safety layer

The bulbs of traffic lights are controlled by the safety layer. It is usually housed in a box next to the intersection.

A microprocessor in this box converts all detector signals and prepares them for transmission. Additionally, this local controller prevents the traffic light from assuming forbidden or unsafe states. Of course, the control algorithm should provide this safety, however the communication could be faulty requiring redundant safety. The components which provide this security are standardised state machines and do not admit logical lapses.

Solid state relays are used for switching the bulbs inside the traffic lights. They convert low voltage control signals to normal 230V on/off signals lighting the bulbs.

An overall view of the system

Figure 1: An overall view of the system

The converter box supervises also the correct operation of all bulbs. In case of bulb failure or short-circuit in one of the cables, the box generates a message automatically and sends it to the process management layer. If a safe operation of the traffic light can no longer be guaranteed in case of loss of one or several critical bulbs, the traffic light flashes amber.

The safety layer only provides safety but no higher-level traffic control functions. These are provided by the process management layer which is connected to the converter box by a simple two wire communication line.

The process management layer

The actual computers controlling the phases of the traffic lights at the intersections are allocated to the process management layer of the traffic control system. They are called node controllers, since an individual intersection is technically speaking a system node. The node controller can be located up to a maximum of ten kilometres away from the intersection. A single current loop (1,2 kbit/s) provides the communication line between the intersection and the node controller. This is relatively slow, but very reliable. It enables the node controller and the intersection logic to exchange detector signals and traffic light commands every 100 ms.

The node controller is the very heart of the traffic control system. It provides an interface to the intersection and handles the communication between node controller and intersection. A single node controller board can handle a single intersection and is part of a multiprocessor-system, which can control up to 90 intersections in parallel.

The control algorithm and the amount of information available (detectors, traffic flows, speeds, behaviour of the traffic participants etc.) determine, of course, the efficiency of the traffic control system.

The node controllers are redundant i.e. each intersection can be controlled by another node controller in case of failure of the original node controller.

The information about traffic light sequences and detectors circulate on a data bus enabling each node controller to read signals from any other intersection in the system. This is very useful, since a single detector can announce the departure of a vehicle from the last intersection and, simultaneously, announce the arrival at the next intersection. This feature increases the degree of flexibility for the control algorithm and allows its design to be independent of hardware configuration and physical locations/connections.

In most cities more than 90 nodes have to be controlled, e.g. in Zurich alone there are around 380 intersections with traffic lights. Therefore the traffic control system consists of several networked traffic control computers. Again, detector signals and traffic light information required for control decisions circulate freely between the traffic control computers. Additional data traffic consists of commands to the traffic control computers themselves and of new or adjusted control algorithms being downloaded into the node controllers.

The traffic management layer

Messages of defective bulbs or messages of malfunctioning control programs are handled by the traffic management layer. The city of Zurich does not present information of intersections on large control panels (with a high entertainment value for visitors), since this is simply not necessary for the supervision of the entire system, neither does Zurich operate video surveillance systems for intersections. Instead, any chosen intersection or group of intersections can be visualised and evaluated on screens in geometry and time in any chosen degree of detail.

On the traffic management layer new control algorithms are developed and tested for controlling existing and new intersections. Testing and evaluation can be done either in the development system itself or later, shortly before the deployment, in the node controller by simulating real traffic which should be controlled optimally.

Additionally, the traffic management layer receives statistical data from the node controllers in order to use it for the evaluation of the traffic flows and for traffic planning. This data is available on-line and can therefore be used for synchronised visualisation.

Back to the intersection: meanwhile the traffic light mentioned before has turned green. Our car may now pass the intersection. Probably, the car will drive over another detector, which will report its departure from the intersection or will announce its arrival to the next intersection.

Had the car driven over this detector during the red phase, it would have announced its departure from the intersection likewise, but possibly it would have provoked a camera to take a very costly picture...

Flows of data

The control algorithms are developed on UNIX-Workstations in the offices of the city police (in the traffic management layer). The draft version is programmed under the object-oriented Modula language (developed at the Swiss Federal Institute of Technology in Zurich). It is tested with simple simulators before deployment, which visualise the phase sequences chosen by the program, and which can simulate triggered detector signals. More refined simulators simulate fictional intersections and feed traffic data into the control algorithm. The control program cannot distinguish between real or fictional data and performs its task independently of the data origin.

The control algorithm is downloaded through the network into the traffic control computer (process management layer) after successful testing where it is assigned to a node controller and starts its operation.

Maintenance and adaptations

The traffic management layer supervises the correct operation of the control algorithms. A control algorithm can be switched off or on at any time or can be replaced by another control algorithm. However, adaptations are always tested in the traffic management layer before they are transmitted to the node controller.

Some data generated by the control algorithms are recorded in files for further use (information on traffic flows/trends, real traffic data test vectors, off-line adaptation of control algorithms).

Implementation in Zurich

The old traffic computers that had been in service for up to 20 years had to be replaced due to insufficient computing power.

Market studies showed that there was no product available on the market which could have adequately replaced the required functions of the existing system. All available computers required massive investments in converter boxes (safety layer), communication equipment and traffic management computers. Zurich therefore decided to develop a modular and open system on its own: the development of a migrateable traffic control system in close co-operation with local engineering and electronics companies.

Step 1: 1991

A PC - supported prototype was developed for controlling the traffic lights of the Heimplatz (a large and fairly complex square in Zurich). The Heimplatz consists of two closely interdependent nodes. Activating and deactivating the prototype was done automatically in order to be able to experiment. The programming language was Modula 2 at the time.

Step 2: 1992

The city police decided to implement a traffic control computer that allowed handling up to 40 nodes, based on the positive results with the prototype. The feasibility of the new traffic control computers was shown by controlling two other complex squares. These were the square in front of Zurich main station (3 intersections) and the combination of Bellevue and Bürkliplatz (5 intersections).

Step 3: 1993 and 1994

The computer was further developed for serial production and the first traffic computers of the old generation were replaced. Additionally, the FDDI backbone was set up which connects the traffic control computers and the traffic management computers (traffic management layer). The modules in Modula 2 were further developed to allow simplified and standardised programming of intersections.

Schedule of the project phases

Figure 2: Schedule of the project phases

Experiences

The implementation proved very successful. Until the end of 1995 the control task for over 200 intersections was transferred to the new system. The replacement of an existing control algorithm consisted of analysing the available control program and drafting, testing, activating and adjusting the new control algorithm.

The switch to the new control algorithm could be done during daytime not requiring night or week-end work. Interruptions of traffic control activity for activating the new traffic control lasted about one minute.

Operation of old and new computers differed only slightly allowing simultaneous operation of the old and the new generation. The replacement of the old generation progressed faster and with fewer problems than originally projected. The availability of the new technology is excellent, and significant reserves in computing power are available in the node controllers.

5. Next steps

Introduction of the new computers has now been largely completed, i.e. the hardware infrastructure is leading edge and up-to-date. Further development of the traffic control system will now be driven by improvements in software and control algorithms.

Simulation

A control algorithm can only be tested realistically using a powerful simulator. Using exact measurements of traffic behaviour, the simulator that has been developed for the new computer system, has to "learn" a wide range of traffic patterns. To this end, the Swiss Federal Institute of Technology has performed extensive research, e.g. resulting in identification algorithms of detector signals enabling the system to reliably recognise specific traffic situations.

Visualisation

The simulator can be used also for the visualisation of an existing control algorithm for any intersection and not only for testing newly developed algorithms. During the visualisation the simulator displays the simulated traffic flows and synchronises simulation and reality by using real detector signals from the intersection.

Algorithms

Detectors around intersections can be inserted to record traffic flows in real time. Two events per detector reading are relevant for the simulation: entering and leaving the detector area. The control algorithm can derive additional information from the time differential between these events e.g. information on speed or congestion. Control algorithms with fairly accurate and real-time information on arriving traffic are superior to fixed-time control algorithms that cannot react adaptively to the current traffic situation at the intersection, especially considering prioritisation of public traffic. The present hardware infrastructure in Zurich ideally allows the deployment of adaptive control algorithms.

Development of control algorithms

The development of the control algorithms is still today largely performed manually by compiling different phase patterns. This work will be automated in future by an expanded simulator enabling also more complicated control strategies, e.g. adaptive control strategies, which have been developed at the Swiss Federal Institute of Technology.

Quality criterion

Until present no commonly accepted criteria for the efficiency/quality of control algorithms have been defined or measured in operation. An efficiency/quality criterion is required to decide whether a control algorithm is sufficiently efficient or whether it still has to be improved.

Co-ordination and precedence

The computer-assisted development of control algorithm enables interlinking and co-ordination of several intersections while still giving public traffic higher priority.