Public Transport Priority Systems: Opportunities and Recommendations

G. Franco and F. Biora (MIZAR Automazione SpA)


CONTENTS

1. EXECUTIVE SUMMARY

2. INTRODUCTION

3. APPLICATION FIELDS FOR BUS PRIORITY

3.1 Fleet control within urban areas

3.2 Selected bus route protection

3.3 Environmental protection schemes

4. POSSIBLE SCENARIOS

4.1 Public vehicles and private traffic interaction

4.1.1 Public vehicles in reserved lanes

4.1.2 Public vehicles in lanes reserved for special vehicles

4.1.3 Public vehicles mixed with private flows

4.2 Priority on a sequence of junctions

4.2.1 Isolated bus route and network of bus routes

5. DYNAMIC ATT METHODS

5.1 Vehicle location/detection

5.2 Adaptive plan signals

5.2.1 Priority recall

5.2.2 Priority extension

5.2.3 Compensation

5.2.4 Inhibit

5.3 Optimised plan signals

5.3.1 Private and public vehicles interaction

5.3.2 Optional stages

5.3.3 Bus stop protection

5.3.4 Signals co-operation

6. ATT VEHICLES LOCATION/DETECTION EQUIPMENT ASPECTS

6.1 On-board location

6.1.1 Use of oedometer only

6.1.2 Use of oedometer plus on-road beacons

6.2 Passage and presence detection

6.2.1 Detectors location

6.2.2 Inductive loops

6.2.3 Transponders

APPENDIX A. THE TELEMATIC SYSTEM FOR THE AUTOMATION OF PT IN TURIN

A.1 Description

A.2 System architecture

A.3 The fleet control feature

A.3.1 Vehicle location

A.3.2 Interaction with the dynamic traffic-signal controller

APPENDIX B. THE SPOT INTERSECTION CONTROLLER

B.1 Description

B.1.1 Optional stages

BIBLIOGRAPHY


1. EXECUTIVE SUMMARY

The bus priority techniques within this document are described in six chapters:

Chapter 2: explains the purpose of the document and how it might be best used.

Chapter 3: lists possible fields, within traffic and travel management and control systems, where bus priority features may improve the effectiveness of the control action. For each field the expected contribution to the bus priority feature is described.

Chapter 4: presents the typical conditions where a bus priority scheme is required to perform an improvement in travel conditions. Problems and characteristics of each scenario are highlighted and solutions are suggested.

Chapter 5: describes the opportunities offered by the adoption of dynamic signal control systems which are equipped for bus priority. An overview of vehicle detection/location methods is also provided as the level of performance achievable with these systems depends on the knowledge of travelling vehicles. This is the core of the document as it deals with advanced telematic systems.

Chapter 6: describes the solutions offered by the current technology for vehicle location/detection to satisfy the requirements of bus priority systems.


2. INTRODUCTION

One objective of bus priority techniques can be improvements in service regularity, which usually means alignment with nominal time-tables and/or headways. A regular service guarantees a good level of transport capacity (expressed in terms of "passengers per hour"): the major goal of transport management. Moreover it makes service planning easier, reduces the time lost by passengers at bus or tram stops, increases user satisfaction and reduces driver stress.

Typical sources of service irregularity are: user demand variations, traffic congestion and traffic signal control. The reduction of the disturbance caused by traffic signal control and the exploitation of priority features constitutes a real success.

A second important objective is a gain in commercial speed. Traffic signal priority contributes to the reduction of PT vehicle journey times and can produce greater transport capacity or a reduction in the number of vehicles required to provide the service.

A third objective, that is becoming increasingly important for transport management, is the reduction of pollution. A smaller number of stops at traffic signals and less time lost in queues are direct effects of traffic signal priority and advanced traffic signal control techniques.

The final important objective is that of a more rational use of energy.

The aim of this document is to report an overview on bus priority techniques as developed and tested in different systems and contexts throughout Europe.


3. APPLICATION FIELDS FOR BUS PRIORITY

3.1 Fleet control within urban areas

Some AVM (Automatic Vehicle Monitoring) systems, such as the one operating in Turin, implement a fleet control action which aims to regularise the public transport service. The regularity of the service is meant as either conformance to the planned timetable or the headway between following vehicles.

One of the main required control actions for such a system is selective bus priority at signalised junctions. This feature allows the system to reduce the journey time of late running vehicles.

3.2 Selected bus route protection

Selected bus routes may be required to provide a service as regular and as quick as possible between any origin and destination. This can be achieved by reserving a bus lane for public vehicles only and providing absolute priority at the signalised junctions.

These measures make the cruise speed of the public vehicles along the route as constant as possible by avoiding any stop other than the planned bus-stops.

3.3 Environmental protection schemes

Environmental protection schemes aim to minimise the levels of pollutants emitted from vehicles. This can be achieved by making the movement of the vehicles within the network as "smooth" as possible, trying to minimise stops and delays at intersections.

For bus priority measures this target can only be achieved if priority is only given to services for which predictable arrivals at the intersections can be provided.


4. POSSIBLE SCENARIOS

4.1 Public vehicles and private traffic interaction

With respect to objectives such as:

  • service regularity

  • gain in commercial speed

  • reduction of pollution

  • rational use of energy

    the best results of the bus priority implementation are expected to be for services operating in protected lanes. Strategies for absolute priority may apply to services operating in reserved lanes, but situations where many services share the same reserved lanes, or where emergency vehicles or taxis can use lanes reserved for public service, have to be handled separately through strategies for high (but not absolute) priority provision.

    For services which operate in mixed lanes with private traffic, strategies which can reduce the number of stops at traffic signals and the time spent in queues are relevant.

    4.1.1 Public vehicles in reserved lanes

    Public transport vehicles operating in protected lanes constitute the class of vehicle which can be given effective priority by any priority system provided with selective priority features.

    The frequency of priority requests is the only major constraint to absolute priority provision.

    This constraint can result in changes to both the traffic signal control and the priority request provider.

    When the intersection is crossed by only one priority service, the frequency should be determined on the basis of the nominal headways of the vehicles on both directions of the service.

    When the intersection is crossed by more than one service, the frequency should be computed on the basis of the headways of all the vehicles which will approach the intersection.

    In any case, if the maximum frequency is not compatible with the average cycle foreseen for the intersection, some restrictions should be applied to the priority request. Reasonable criteria are:

    a) Reduce the level of priority for one or more services from "absolute" to "high".

    This rule can be applied only if services do not share the same route. Otherwise lower priority vehicles could obstruct the progress of the absolute priority vehicles.

    b) Request priority only for vehicles running late.

    This rule could be applied either to one, to a subset, or to all of the priority services. Some services could maintain the absolute priority if the resulting request frequency becomes comparable with the average length of the traffic signal cycle. The other vehicles could be provided with a "high" level priority.

    As a consequence of specification a) defined above, possible conflicts between priority requests of the same level or of different levels should be solved by the intersection controller.

    4.1.2 Public vehicles in lanes reserved for special vehicles

    A "reserved lane" is defined here as a non-protected lane which is reserved for public transport vehicles. The access to a reserved lane is forbidden to private traffic but can be permitted to vehicles of other public service providers (emergency vehicle, police, taxis and so on). Normally reserved lanes are shared by many services (trams and buses).

    For this class of vehicles journey time predictions present higher variances. The intersection control can still provide a high level of priority.

    Priority assignment at the intersections can still be constrained by the priority request frequency. Problems due to possible obstacles to the vehicle's progress can occur.

    Specifications a) and b) defined in 4.1.1 can apply also to this context only in some particular cases: reserved lanes shared by a few services, all provided with the same level of priority and not disturbed by vehicles with lower priority.

    In other cases a common level of priority (lower than absolute) should be defined for the set of services sharing the reserved lane. The evaluation of the request frequency could become really complex: conflicts with other services using different reserved lanes can be present in several intersections. A general specification is:

  • to request priority only for late vehicles, and/or

  • to solve conflicting requests at the intersection level.

    4.1.3 Public vehicles mixed with private flows

    The importance attributed to the single service should lead to the definition of the weight of the priority requests for different vehicles.

    Also for this class of vehicles the accuracy of journey time prediction is sufficient for the formulation of priority requests, but the priority assignment depends strongly on special features of the intersection control. Indeed, in this context the real problem concerns private traffic which can obstruct vehicle motion towards the downstream intersection.

    The intersection control should be provided with features for lane clearance. The clearance action should start with the first vehicle arrival time forecast received by the intersection controller, and continue until the vehicle has left the intersection.

    This technique could be implemented by increasing the weight of the incoming link used by the vehicle, and consequently by influencing the functional optimisation performed by the intersection controller.

    This kind of action in general can introduce a disturbance in the traffic control. Hence, priority requests for vehicles mixed with the private traffic should be limited.

    Priority request for vehicles mixed with private traffic should be limited to extremely late vehicles.

    The great number of possible priority requests which can be generated at network level suggests a general specification: the weight of the priority requests should also be defined according to the variance of the predictions of the arrival time of the vehicles.

    The intersection controller should be able to manage priority request weights defined dynamically and to solve possible conflicts between requests accordingly.

    4.2 Priority on a sequence of junctions

    The previous section analyses the possible scenarios for bus priority at a signalised junction according to different topological layouts of the intersection. General specifications and considerations have been formulated both for the local level controller and the priority requester about the priority level to be requested and the accuracy of the arrival time prediction.

    These considerations have to be integrated by others when priority is requested in scenarios where intersections are not isolated but are part of a controlled network.

    The simplest and most frequent of these cases may be represented by a sequence of controlled junctions on a corridor.

    In this scenario an information flow between the local controllers, or even from the local controller towards the system that requests priority, is needed in order to take into account the delays introduced by the sequence of controlled intersections crossed by the public vehicles.

    The information exchanged should include the information about the predicted arrival time of the vehicle at the first junction that will be crossed, the journey times between controlled intersections and the additional delays introduced by the junctions that will be crossed.

    Looking at the network as a whole, the arrival forecast envisaged in the previous section should become a predicted time trajectory along the corridor or the network depending on:

  • the route of the vehicle

  • the priority level requested for the vehicle

  • the current traffic conditions of the crossed intersections

    The interaction between the systems becomes much more complicated as the scenario changes from a isolated bus route running on a corridor to a network of bus routes running on a controlled network.

    4.2.1 Isolated bus route and network of bus routes

    In the case of a bus route running on a corridor or on a sequence of controlled junctions the interaction between intersections can be solved at both the local controller level and at the priority requester level.

    Under these conditions identification of the vehicle is not needed because all the vehicles follow the same path.

    Interaction could then be resolved at the local level. This should allow a good estimate of the delays introduced by the junctions on the vehicle's route to be estimated.

    On the other hand this approach requires the local controller to be capable of estimating journey times between the intersections.

    As the journey time estimation may become complicated if the route includes bus-stops this solution can only be envisaged when the topology of the route allows the use of fixed journey times.

    The presence of bus stops or a complex topology of the route would need the interaction to be resolved at the priority requester level.

    With this approach the prediction of the delays on the route becomes more difficult, especially when the intersection control goes towards full decentralisation, but journey times can be estimated with a much greater accuracy.

    When priority is intended to be given to vehicles running different routes on a network the problem of resolving conflicts at each intersection between different routes and deciding the priority level for each route arises.

    Identification of the vehicles is an essential point of the above mentioned problem and this leads to the separation of the local controller functions from the priority requester.

    As stated in section 4.1 these problems have to be solved at the priority requester level while the local level controller should be able to manage priority requests with different levels of priority. Again an interaction between the traffic signal control and the priority requester may be envisaged in order to estimate the delays to the public vehicles in the network of controlled junctions.


    5. DYNAMIC ATT METHODS

    In the ATT environment, dynamic PT priority is usually considered as a function of an UTC (Urban Traffic Control) system; these systems are usually based on the identification of a "minimum cost" control policy; different elements are combined in order to determine the "cost" of a policy: private vehicles interest (minimisation of time lost in queue and/or of travel time), public vehicles needs (priority, regularity of service), environmental constraints (minimisation of queues and stops in order to reduce pollution, avoidance of congestion).

    The PT traffic is considered as a component of the global traffic, with peculiar features and needs; for this reason, most of the UTC systems aim to optimise the overall performance of the road network, introducing the PT priority as a constraint to the optimisation problem, separating as little as possible the problem of PT priority from that of private traffic control.

    A real-time UTC system is able to react in a short time interval, to each perturbation in the traffic, typically modifying the setting of the traffic signals at an intersection to allow a bus to pass without stopping (in this case, the arrival of a PT vehicle is seen as a "perturbation"). In this approach, the system must be able to perform the following tasks:

    recognise the PT vehicle rapidly and accurately, to meet the needs of the control strategy,

    implement an effective and rapid optimisation of the local policy, in order to minimise the disadvantages for the private traffic (e.g., stopped vehicles in the crossing direction when a bus is given the priority),

    perform the procedures involved in the priority task with suitable speed.

    5.1 Vehicle location/detection

    Dynamic ATT approaches for bus priority require the availability of a vehicle detection system which allows on-line forecasting of approaching PT vehicles.

    The PT detection system can be implemented in three basic ways:

    Inductive loops on the reserved lane; this technique gives either the PT vehicle presence or passage over the loop location. The detection cannot be vehicle specific so the detectors must be located on reserved areas to avoid confusion with other vehicles. Configurations of loops can be provided for each junction aiming to follow the movement of the vehicle within the signalised intersection area.

    Beacon or transponder; this technique allows a vehicle specific detection and a limited information exchange such as level of required priority and forecast arrival time. The PT reserved lane is not necessary but the information is constrained to fixed locations.

    UTC-AVM integration; the arrival forecast of the public vehicles is provided via AVM which can usually provide information updating with variable frequency which is not constrained to fixed locations.

    The performance achievable by the UTC system is dependent on the kind of detection available as well as on the accuracy of the information provided by the detection system.

    5.2 Adaptive plan signals

    This approach is based on the "green call" concept: a detected PT vehicle which is approaching an intersection activates a request for the green; the local controller reacts, subject to several constraints, either by the extension of the current green stage or by the early activation of a suitable stage (the suitable stage can be either reserved for public vehicles only or for mixed traffic).

    The basic functions which have to be performed by the local controller are described in the following subsections:

    5.2.1 Priority recall

    When a bus arrives during a red stage, a suitable green stage is run as soon as possible, with the constraint that competing stages must be run in the usual order, although for the minimum length.

    This is the main function which characterises the priority approach, the effectiveness of this function depends on the accuracy of the forecast arrival time of the approaching PT vehicle.

    5.2.2 Priority extension

    When a bus approaches the detector while a green stage is actuated, the stage can be extended in order to allow the bus to pass undelayed. The extension is computed using the free-running time of the bus between the detector and the stop line. In the case of more than one arrival on the same stage, the extensions can be accumulated up to a fixed maximum.

    This function is complementary to the function described in 5.2.1 which allows the exploitation of suitable stages which could be provided within the standard cycle.

    5.2.3 Compensation

    When a cycle is modified by a priority procedure, the following cycle is modified too, in order to recover the time lost by the non-priority flows. The time added to the stages of the cycle can either be fixed for each intersection or reckoned dynamically based on the traffic demand.

    This function works as a complement to the "green-call" functions with the aim of balancing the effects of the priority on the private flows.

    5.2.4 Inhibit

    In presence of heavy PT flows, introducing the compensation for non-priority stages, as described in 5.2.3, can be difficult due to the sequence of the calls. Moreover, in some cases full priority is considered unacceptably disruptive for the traffic flow. In these cases, priority is permitted in alternate cycles only.

    5.3 Optimised plan signals

    ATT systems which adopt on-line optimisation of signal plans are based on forecasts of traffic conditions. The traffic flow estimate is reckoned and updated by processing the data provided by the on-road detectors.

    Traffic control strategies are dynamically planned in advance in order to match both the private traffic and PT estimated demand.

    The strategy optimisation is focused on the choice of the green split and cycle time for each intersection as well as on the synchronisation between adjacent signalised junctions.

    5.3.1 Private and public vehicles interaction

    The representation of the public vehicles with respect to the private vehicles can be different as it depends on the system used.

    Some of the commonly used representations are:

  • PT vehicles are modelled independently as dynamic constraints for the optimisation algorithm rather than particular vehicles which move within the network. This approach is related to the "green-call" concept explained in 5.2 which could also be use within this context.

  • PT vehicles are modelled as a continuous group of cars which moves within the network following the same model used for the private cars. This approach is an attempt to model the number of moving passengers rather than vehicles. For different models, the number of cars which are included in the group, can be chosen as either:

  • fixed for every PT vehicle,

  • fixed and related to the service route,

  • reflecting the delay of the PT,

  • reflecting the number of passengers carried

    or any mix of the above criteria.

    Whatever is the representation of the PT vehicles, the balance between public and private vehicles can be tuned in order to reach the requested "level of priority" (i.e.: the priority is absolute if the rate is set as 1 public vehicle = 500 private vehicles).

    When public and private vehicles share the carriageway their interaction must be taken into account. A PT vehicle has to wait for the queue of private vehicles to be cleared before it can cross the intersection. This effect can be taken into account by systems which are able to dynamically estimate the traffic conditions.

    5.3.2 Optional stages

    Optional stages are stages which are normally not included within the cycle but can be suitably planned whenever a PT vehicle approaches possessing specified conditions.

    For instance, this feature is used when a PT vehicle can benefit from turning movements which are not available for private traffic.

    These stages can be planned within the cycle either by matching them with PT arrival forecasts or as a response to stage calls from on-road detectors.

    The length of an optional stage is usually variable, within fixed limits, so it is possible to overcome problems such as forecast errors and calls which are close to each other.

    When there is a call for an optional stage, the optimisation algorithm calculates a length for the optional stage length which balances the traffic demand against the cost for a waiting PT vehicle.

    Typical basic scenarios are:

  • Forecast of approaching PT.

    The optional stage is activated when the vehicle is expected to be close to the intersection. If the forecast is accurate enough the stage can be actuated until it has been confirmed that the vehicle has been released.

    Vehicle specific features are possible if information exchange is available.

  • Approaching PT detection.

    A single approaching detector provides a fixed approaching time and the stage has to be actuated for a fixed time long enough to make sure the vehicle manages to cross the junction. Because the released vehicle is not controlled directly, stage time is wasted in order to ensure the effective PT passage.

    Vehicle specific priority features can be implemented if the detectors are transponders.

  • Presence detection.

    A detector placed nearby the stop line is designed to detect the presence of a waiting vehicle. The optional stage can be extended, within fixed limits, if the detector is longer occupied. This kind of detector provides information that a vehicle has been released into the downstream link if the crossing time is assumed (The vehicle is assumed released after the crossing time following the end of the detector occupancy).

    Vehicle specific priority features can be implemented if the detectors are transponders.

  • Approaching and exit detection

    A set of detectors completes the features available adopting the approach only detector with the released vehicle information.

    Complex schemes, such as different bus routes which require different optional stages for different turning movements, can be implemented.

    Combinations of basic scenarios are also possible aimed at improving the effectiveness of the detection system:

    forecast aided by on-road detection

    approaching plus presence plus exit detector layouts

    are two examples.

    5.3.3 Bus stop protection

    When the carriageway is shared a queue of private vehicles can obstruct the approach to an upstream bus stop causing a delay in the PT journey. In order to remove this delay the control system should manage to reduce the occupancy of an approach to the intersection when the bus is approaching a stop in the neighbourhood of the junction.

    5.3.4 Signals co-operation

    Control systems which optimise on long prediction horizons can improve their PT priority if the release time of public vehicles from up-stream junctions is communicated to the down-stream junctions. This technique overcomes a possible lack of detectors and provides an approaching PT forecast with long notice.

    When this technique is used the on-road detector information provides the approaching forecast update. A time window can be created in which the vehicle is expected to run over the detector:

  • if the vehicle is effectively detected within that window then the actual detection time is used to update the estimated arrival time,

  • if the detection is not confirmed within the interval the expected vehicle is declared "lost" and deleted from the approaching vehicles list; it is included again as soon as it is detected.


    6. ATT VEHICLES LOCATION/DETECTION EQUIPMENT ASPECTS

    In this section an overview of the most commonly used equipment for vehicle detection is reported.

    The description aims to highlight the peculiarities of the techniques describing broadly the features and without specifying technological details which depend on the specific applications and hardware used.

    6.1 On-board location

    On-board location is one of the methods that can be used to dynamically provide the position of vehicles within the network and supply arrival time information.

    The technique applies both to centralised and decentralised architecture. The vehicles are periodically polled by the system which needs to know the current position and to work out the forecast arrival time.

    The most advanced techniques make journey time predictions by observing historical as well as current vehicle presence times within defined parts of the bus routes. Improvements of the approaching PT prediction are achieved by varying the frequency of enquiries of the vehicle position and consequent forecast data update: the closer the vehicle is to the controlled junction the more often the vehicle position is sampled.

    6.1.1 Use of oedometer only

    One successful technique uses an oedometer connected to the wheel of the vehicle, the counting being processed by an on-board computer which also contains the vehicle route map. The computer keeps track of the position of the vehicle within the route and updates the information, recognising the occurrence of the bus-stops by observing the activation of the doors of the vehicle.

    Identification algorithms are provided in order to overcome possible skipped bus-stops and slight deviations from the theoretical route.

    The vehicle position is reset by the driver when the vehicle reaches the terminal.

    This method makes the vehicle location information available at any time by contacting the vehicle.

    6.1.2 Use of oedometer plus on-road beacons

    The same vehicle location technique described in 6.1.1 can be used in an architecture which also includes on-road beacons. The knowledge of the position of the beacons within the route map is used by the on-board computer, either in addition to the door activation or not, to update the estimate of the vehicle location.

    6.2 Passage and presence detection

    When a vehicle location system integrated with the UTC system is not available, the on-road detectors are the only source of information for the intersection controllers designed to provide PT priority.

    6.2.1 Detectors location

    There are a wide range of suitable locations for detectors of public vehicles. For bus priority purposes they have to provide information which is as reliable as possible about PT vehicles which are about to approach controlled junctions.

    Dynamic strategies require an approaching detector as far as possible from the stop line so the information can be provided with long notice.

    The best location is usually at the entry to a link, just downstream of the upstream intersection so that the uncertainty introduced by the possible waiting time at that junction can be avoided.

    If there is a bus-stop on the approach link the best detector location is just downstream of the bus stop bay so the uncertainty due to the stop time can be avoided.

    Presence detectors have to be placed as close as possible to the stop-line so they can also provide the released vehicle information (see 5.3.2). Presence detectors are not suitable when a PT reserved lane is not provided as the queue of private cars could keep the PT vehicles waiting outside the detection area.

    When the public vehicle reaches the junction along a mixed traffic carriageway and it then needs to call an optional stage in order to make a turn which is in conflict with private traffic, a presence detector beyond the stop line may be used, located in the junction where the PT vehicle waits to turn.

    For this situation, in some cases the change point activation, operated by the PT vehicle, can be used in addition to the presence signal.

    As has been stressed in 5.3.2, often exit detectors are useful in order to confirm the effective passage of the PT through the intersection. For this purpose the most suitable position for the detectors is downstream of the PT turning where the vehicle completely clears the intersection.

    6.2.2 Inductive loops

    The inductive loop is probably the most widespread technology for vehicle detection. It can be used for the particular case of PT vehicle detection. Loops placed within PT reserved lanes have the same requirements as those designed for detecting private vehicles.

    Particular design and tuning must be adopted where the loops are placed in carriageways shared with the private and public vehicles in order to avoid unwanted wrong detection.

    The difference of mass between private and public vehicle is usually the parameter exploited to distinguish between them.

    Particular care must be taken in placing the PT detectors if there are likely to be problems with masking due to parked vehicles. Illegal parking has to be considered too.

    6.2.3 Transponders

    These detection devices allow the detection of the equipped vehicles as well as a limited exchange of static information which could be optionally stored (i.e.: bus route number, level of priority requested).

    These systems provide the equipped vehicle with a passive communication unit (transponder) which transmits the stored information when the vehicle runs over the on-road detectors. In this case the on-road detectors are aerials which receive the data while transmitting the power supply.

    The use of such a system allows the identification of specific vehicles and consequently the implementation of features of vehicle specific priority. Furthermore they remove problems due to carriageway sharing with other vehicles, which now do not benefit from the priority as only equipped vehicles can be detected.


    APPENDIX A. THE TELEMATIC SYSTEM FOR THE AUTOMATION OF PT IN TURIN

    A broad description of the system is provided in this appendix with particular emphasis on the bus priority feature. Bus priority at junctions is available in this system as actions of a PT fleet control system for the regularisation of the service.

    A.1 Description

    The system which is operating in Turin is a telematic system designed to assure assistance to public transport with a centralised fleet management architecture.

    The main features available are:

  • automatic regularisation of the PT service

  • monitoring of the PT fleet

  • data and voice radio link between vehicles and control centre

    The objective of the system is the control of disruptions and perturbations of the PT service aiming to produce a service which is as punctual and regular as possible.

    This result is achieved by the integration of the system with the traffic-signal network control system and the customer information system.

    A.2 System architecture

    The architecture is based on a mobile radio system which allows the transmission of data and voice between control centre and public vehicles.

    The main components of the system are:

  • the control centre, which is equipped with a central computer dedicated to the communication management, the control and regularisation features and the operator interface.

  • the telecommunication system, which implements the data and voice radio link between vehicles and centre

  • the on-board equipment, which keeps track and transmits data on request about the current location of the vehicle and the number of passengers. The phone equipment for the voice communication between drivers and centre is also provided.

    A.3 The fleet control feature

    The control action of the PT fleet is based on two main functions:

    vehicle location and journey time processing

    service regularisation

    A.3.1 Vehicle location

    This function is performed in an integrated way by the on-board and the central systems.

    The on-board system deals with the estimation of the current position of the vehicle along the route using the data provided by the oedometer which is installed on the vehicle wheel. The vehicle position is transmitted to the centre on request.

    The aims of the location software implemented on the central computer are:

  • validation of the data provided by the on-board system

  • identification of the actual position of the vehicles on the network as the information provided by the vehicles themselves is not complete

  • vehicle polling in order to get the location information in sufficient time for the forecast progression

  • journey time processing aimed at compiling a dynamic timetable and predicting the progression of the vehicles on the network.

    The location estimation is performed with statistical filtering; the estimation is reckoned by comparing the measurements of the oedometer (2m resolution) and the route description as stored within an EPROM. The vehicle door activation at the bus-stops is used as additional information. The standard deviation of the location error achieved by this technique is <5m.

    A.3.2 Interaction with the dynamic traffic-signal controller

    The interaction of the system with a centralised traffic-signal controller is provided (green stage call) as the dynamic priority of PT at the signalised junctions is a basic requirement for fleet irregularity control.

    The traffic-signal network controller used in Turin is the UTOPIA system which includes local SPOT type controllers which are co-ordinated by a traffic control centre (see Appendix B).

    Important aspects of the priority at the signalised junctions are:

  • it is not feasible, in a grid network, to assure absolute priority to all the running PT vehicles as they can be in conflict with each other. Therefore a priority order between different vehicles has to be defined. That priority order must be dynamic in order to match the regularisation algorithm requirements

  • the priority at signalised junctions requires that the approaching vehicle information is provided to the single junctions with sufficient notice and accuracy.

    The functions performed by the integrated system are the following:

  • dynamic priority order reckoning of the vehicles according to the regularisation algorithm needs and road characteristics

  • communication to the traffic-signal controller of the PT approaching time as well as the level of priority required for the vehicle

  • interaction with the vehicle location module for the intersection approaching time forecast.

    The accuracy available for the vehicle approaching time depends on the forecast horizon: it is <10% for a horizon of about 5 minutes, but up to 30% for a horizon as long as 30 minutes.

    The approaching vehicle information is sent to the UTC system as soon as the vehicle is within two minutes (the horizon of the intersection control optimiser) of the intersections. Forecasts are then updated according to the progress of the vehicles. For this purpose the central system, which usually samples the vehicle location once a minute, queries the vehicle before the intersection up to every fifteen to twenty seconds until the vehicle has left the junction according to the latest forecast.

    Vehicle delays, for the determination of the level of priority needed, are computed by the central system by comparing single vehicle positions to a 'virtual schedule' (computed on-line on the basis of nominal schedules and headways). In order to forecast in advance service irregularity, vehicle delays are computed on the basis of forecasts of their arrival at advanced targets put on the route.


    APPENDIX B. THE SPOT INTERSECTION CONTROLLER

    This section contains a description of the basic concepts of the SPOT intersection controller which is a UTC unit designed to optimise the signal settings of a single intersection, an isolated small network of controlled intersections as well as a network supervised by a central system.

    B.1 Description

    The control function which operates at the intersection level determines the signal settings to be applied to the traffic signals by optimising a suitable function according to the current intersection traffic situation. The optimisation is done on a 'time horizon' of the next 120 seconds and is repeated every three seconds. The resulting optimal signal settings are actually in operation only for three seconds. The closed loop control thus obtained can be viewed as an 'Open Loop Feedback Control' or as an application of a 'Rolling Horizon' concept.

    In order to guarantee the optimality and the robustness of the control at the network level, the function optimised by the controller has been designed by adopting the 'strong interaction' concept: the function takes into account the state of the neighbour intersections, thus keeping a closed loop capability of building a dynamic signal coordination, and is constrained by limits given by the area level control (while remaining sensitive to traffic dependent criteria).

    The function is defined by the sum of different weighted cost elements calculated on the whole optimisation horizon. The optimisation goal is:

    equation

    subject to constraints such as:

    equation

    where

  • wj = weight of the cost element 'j'

  • aj = cost element 'j'

  • c = signal setting

  • x = intersection state defined on the whole optimisation horizon

  • u = intersection demand

    condition (2) represents the constraints on the length of the stages.

    The cost elements are:

    1. Time lost by vehicles on the incoming links.

    2. Stops on the incoming links. Stops are defined as vehicles which arrive at the stop line when queues are present.

    3. Excess queuing on the incoming links (this term denotes queues exceeding safety thresholds which are proportional to the maximum capacity of the links).

    4. Time lost on the outgoing links by vehicles leaving the intersection (these terms actuate the 'strong interaction' principle at the intersection level. They provide intersection control coordination and control stability at the area level).

    5. Time lost by public transport vehicles to be given priority at the intersection.

    6. Deviation from the reference plan provided by the central level (this element actuates 'strong interaction' with the area level and allows the degree of interaction between the two levels to be dynamically changed).

    7. Deviation from the signal setting decided at the previous iteration (this element contributes to the smoothness of the area control)

    Cost elements are evaluated on the whole horizon on the basis of traffic propagation rules which take into account the signal settings and the constraints on the minimum and maximum stage lengths. Different weights are allowed for different links and for different PT services (for providing absolute and lower level priority).

    Traffic propagation at the intersection starts from the state estimate provided by the observer and makes use of all the defined traffic parameters. Input information required is as follows:

    For the incoming links:

    1. Traffic counts

    2. Traffic forecasts provided by the neighbouring controllers (Forecasts correspond to vehicles which will leave the upstream intersections. An approximation is made assuming that the outgoing flows are uniform at intervals).

    3. Forecasts concerning the arrival of the public transport vehicles to be given priority.

    For the outgoing links:

    4. The control strategies defined by the downstream controllers

    Public transport vehicles are represented by equivalent vehicle platoons which appear as probability curves centred on the predicted arrival times. Curves becomes sharper as the vehicles approach the intersections and forecast variances decrease. The weight of the single vehicle depends on the level of priority requested. Weights can assume values within a suitable range defined on the basis of a sensitivity analysis of function (1).

    Currently, absolute priority vehicles correspond to four-five hundred equivalent vehicles and the weight of normal priority vehicles depends on the weight predefined for the corresponding services.

    When no PT vehicle priority is requested the intersection controller provides the optimal traffic signal control according to the private traffic conditions only. The intersection controller is then able to satisfy priority requests even without any significant disturbance of private traffic through:

  • Gradual adjustments of the traffic signal stages (in terms of duration and actuation time).

  • Gradual adjustments of the synchronisation with the neighbouring controllers.

  • Actuation of stages whose duration is as close as possible to the duration suitable for private traffic control (according to the weights introduced in the function optimised).

    The intersection control performance depends on both the availability of PT vehicle arrival time forecasts updated in time according to the vehicle's progress, and the number of requests to be solved together. The experimentation performed demonstrated that the priority system was able to provide absolute priority to PT vehicles approaching the intersection once per cycle (in protected lanes) and simultaneously to optimise private traffic control.

    B.1.1 Optional stages

    Optional stages are included within the signal cycle during the horizon optimisation, whenever waiting or approaching public vehicles are forecast on suitable links.

    Seeking the optimum strategy, for each horizon step the possible stage is reckoned according to the following algorithm:

    First optimisation step (long horizon)

    The optional stages are "skippable" if they don't improve the optimisation function cost which, at this level, takes into account strategy variation and bus priority costs only.

    Second optimisation step (short horizon)

    a) Select the stage following the previous step stage within the whole cycle (including all the optional stages)

    a1) If the selected stage is not optional, it is the only possible following stage.

    a2) If the selected stage is optional, the presence of queued vehicles and the presence of forecast approaching vehicles on the proper link are evaluated. If the evaluations show the stage is needed, it is the possible following stage else the test is repeated selecting the next stage within the whole cycle.

    Approaching PT vehicle forecasts, in the form of probability distributions, can also be carried out according to information provided by special detectors installed on-road.

    Two methods for bus priority are currently managed:

    Presence detection

    One stopped waiting vehicle is generated on the appropriate link while the loop is occupied.

    Single approaching vehicle detector

    Every time a variation on-off is received from the loop, one approaching vehicle forecast is produced on the appropriate link (the detected vehicle delay is selectable).

    Following vehicles will pile up at the stop line until the beginning of the favourable optional stage. During the optional stage, queued vehicles are released at the rate of one vehicle per step.


    BIBLIOGRAPHY

    Lanteri, F. (1992), "Features of Public Transport Priority Systems", DRIVE II Project V2016 : PRIMAVERA, Deliverable No. 2.

    Lanteri, F., Biora, F. and Shepherd, S. (1993), "Implementation Aspects Report", DRIVE II Project V2016 : PRIMAVERA, Deliverable No. 8.

    Davidsson, F. (1992), "Interim Report on Requirements for Public Transport and Informatics", DRIVE II Project V2049 : PROMPT, Deliverable No. 4.

    Mauro, V. (1991), "Road Network Control", Concise Encyclopedia of Traffic & Transportation Systems, Pergamon Press, Oxford.

    Mauro, V. and Di Taranto, C. (1989), "UTOPIA", CCCT '89 AFCET Proceedings September 1989, Paris.

    Gentile, P. and Mauro, V. (1988), "Experience with SIS, Torino's Public Transport Operation Aid System", International Conference on Automatic Vehicle Location in Urban Transit Systems, Ottawa.

    de Saint Laurent, B. (1991), "An Information System for Public Transport", Proceedings of the DRIVE Conference, February 1991.


    ITS Home PageITS home page