by Ken Fox[I], Frank Montgomery[I], Simon Shepherd[I], Chris Smith[P], Simon Jones[H] and Fabrizio Biora[M]
[I] Institute for Transport Studies - University of Leeds, Leeds. LS2 9JT,
[P] PEEK Traffic Ltd, Sovereign House, Stockport Road, Cheadle, Cheshire. SK8 2EA,
[H] WYHETS, SELECTAPOST 6, Dudley House, 133 Albion Street, Leeds. LS2 8JX,
[M] MIZAR Automazione SpA, Via Vincenzo Monti 48, Torino, 10126, Italy.
Introduction
This is an article about DRIVE II project PRIMAVERA[1]. It describes a cost effective bus priority system using TIRIS transponders that was developed by the project. This detection system has been fully integrated with both the SCOOT[2] and SPOT UTC[3] systems. In field trials in Leeds this resulted in an 8% reduction in travel times during the morning peak for those buses fitted with the transponders.
TIRIS
Introduction. The Texas Instruments Registration and Identification System (TIRIS) is a radio frequency identification (RFID) system based on low frequency FM transmission techniques.
The core of the system is a small transponder or tag that can be attached to or embedded in an object. To interrogate the tag, a reader sends out a radio signal to the transponder via an antenna. The signal carries enough energy to charge up the passive (battery free) transponder. The transponder then returns a signal that carries the data that it is storing. This data is a unique, factory programmed identifier.
How it works. The three major parts of the TIRIS system are the transponder, antenna and reader. When a transponder is to be read, the reader sends out a 134.2 kHz power pulse to the antenna lasting typically 50 milliseconds. The field generated is collected by the antenna in the transponder that is tuned to the same frequency. This received AC energy is rectified and then stored in a capacitor within the transponder. When the power pulse has ceased, the transponder immediately transmits back its data, using the energy stored within its capacitor as the power source. A 64 bit identification together with error detection information is transmitted over a period of 20 milli-seconds. Once all the data has been sent the storage capacitor is discharged, thereby resetting the transponder ready for the next read cycle. The data is picked up by the antenna and decoded by the reader unit. The transmission technique used between the transponder and the reader is Frequency Shift Keying (FSK) using 134.2 kHz and 123.2 kHz. This approach has good resistance to broadband noise whilst being very cost effective to implement.
How it is used to detect buses
Introduction. The TIRIS system was originally developed
to keep track of livestock, however it was soon realised that there was
a great potential in many other application areas. One of the areas developed
was automatic vehicle identification. DRIVE II project PRIMAVERA was an
EU funded international collaborative project involving Leeds City Council,
PEEK Traffic Ltd, MIZAR Automazione SpA, Turin City Council, with the Institute
for Transport Studies, University of Leeds as project co-ordinator. PRIMAVERA
focused on integrated traffic control on urban arterial corridors. It developed
and appraised Advanced Transport Telematics (ATT) methods for managing
queues of traffic, giving priority to public transport, and improving safety
through traffic calming measures[1]. These techniques
are potentially conflicting in their effects. The project developed means
of integrating the individual control measures to ensure that they complemented
one another and obtained an acceptable balance between objectives. The
consortium decided to develop and install a low cost bus priority system
based on TIRIS transponders as part of the field trials in Leeds. Leeds
City Council obtained funding and installed the necessary loop detectors
and TIRIS readers at four junctions on the Dewsbury Road trial site. The
main bus operators purchased and installed the TIRIS transponders on their
fleets of buses at their own expense. The messages produced by this system
have been integrated with both the UK SCOOT UTC system and the Italian
SPOT UTC system. Both systems have been tried out in field trials in Leeds
during 1994.
Figure 1: TIRIS Unit
The TIRIS transponders are cylindrical in shape, approximately 25mm in diameter and 125 mm long including plastic surround. They are attached to the bus chassis on the underside of the bus. The transponder is mounted vertically within a moulded plastic bracket that is designed to place the transponder 100 mm away from the metal surface to which it is attached. (See Figure 1) Provided the transponder is within 1.5 m of the road surface the location on the chassis is not critical to its operation. The system is very robust to noise and the manufacturers claim that it has successfully operated at vehicle speeds up to 150 km/h.
A system was developed by JMW Systems Ltd, under contract to Leeds City Council Department of Highways and Transportation, to provide bus detection information, using TIRIS back to a UTC system.
The TIRIS reader is mounted on a steel chassis inside a sealed enclosure on a pedestal at the side of the road. This drives an antenna embedded in the road. The antenna consists of a rectangular loop of cable. The cable used was the same as for vehicle detector loops at traffic signals. Antennas were installed both on two way roads and divided carriageways to detect a single direction of flow. The road half-widths varied between 4 and 8 metres. The loop perimeters were similar in all cases, approximately 15 to 16 metres. Consequently the length of the loop in the direction of travel varied from 1 to 5 metres. Whereas loops for urban traffic control are placed at the entrance to links, the loops for bus detection are best placed between a bus stop and the junction stopline. This allows a good prediction of the arrival time of the bus at the junction to be made.
Error checking. It is essential that the system only detects buses fitted with the transponders and not any other vehicles. (E.g. some cars are now fitted with TIRIS transponders as a security measure) For this reason an interface card, within the reader enclosure, stores the valid range of transponder identification numbers for comparison with those received. Should the number not be valid it is ignored.
A problem that has been encountered with early bus detection systems has been multiple detections of the same vehicle causing excessive demand for priority. The TIRIS system overcomes this due to the unique identifier associated with each transponder. On detection of a valid number a pair of isolated relay contacts are opened for 4 seconds, so that this particular number will now not reactivate the system for five minutes. The system will also not activate the relay contacts for a period of 0.5 seconds after which it will start looking for a new transponder. Should no valid transponders be detected within a pre-set period of between 18 and 72 hours an alarm lamp is illuminated and the system is switched off.
Connection to the UTC system. The TIRIS readers are equipped with simple relay outputs which change state after the detection of a bus for which priority is to be requested. These outputs are either connected via cables and buffering relays to the nearest SPOT microprocessor or, in the case of SCOOT, to the nearest Outstation Transmission Unit and hence back to SCOOT.
PEEK Traffic modified the SCOOT software to allow the system to give priority to the detected buses. Likewise, MIZAR Automazione modified the SPOT software.
Cost. A single TIRIS transponder costs approximately £25. A TIRIS reader unit costs approximately £1500 to install at the roadside. A TIRIS loop antenna costs approximately £200 to embed in the road and connect up to the reader unit. This TIRIS transponder based system therefore provides a highly cost effective way of giving priority to public transport.
Implementation for the PRIMAVERA field trials. In Leeds, strategies
have been developed which are implemented using either the SCOOT UTC system
or the SPOT UTC system, that are used to control ten signalised intersections
along the Dewsbury Road. Four of these junctions were equipped with TIRIS
readers to detect buses travelling towards the centre of Leeds. (See Figure
2). The Dewsbury Road, classified as the A653, is one of the main radial
routes into Leeds. It serves as a main commuter route to and from the towns
of Dewsbury, Morley and parts of Wakefield as well as the M62 Trans-Pennine
Motorway and the local communities of Beeston and Middleton.
Figure 2: Passing information from TIRIS to the UTC system
The 3km length of Dewsbury Road chosen for the field trial extends from the A6110 Leeds Outer Ring Road to its intersection with the M1 Leeds to London Motorway. It carries, on average, a traffic flow of approximately 23,000 vehicles per day.
Dewsbury Road is a heavily used public transport corridor. The inbound bus flow (e.g. towards the city centre) is 38 buses in the morning peak hour (8.00 to 9.00), while 36 buses travel in an outbound direction in the evening peak (17.00 to 18.00).
There are 24 bus stops on the stretch of Dewsbury Road chosen for the trial, which are sited at irregular intervals. The stops mainly serve commuters and shoppers on their journeys into and from the City Centre, but some have a local function serving shops and businesses in the Dewsbury Road area.
PRIMAVERA is concerned with integrating queue management, public transport priority and traffic calming strategies
Two public transport priority strategies have been investigated for application at the Leeds trial site.
* Bus progression (P)
This strategy is implemented using the SCOOT UTC system.
In common with most UTC systems, SCOOT tries to optimise the traffic signal settings to benefit the traffic stream as a whole. Traffic does not, however, behave in a homogeneous manner. In particular a set of signal offsets geared to providing a green-wave to traffic as a whole tends to provide a red-wave to buses. It has been shown[4] that in a fixed time context, significant benefits can be achieved if the bus traffic stream is separated out and given extra weight.
Within the context of PRIMAVERA, a similar benefit can be achieved by placing an additional constraint on the offset optimiser. A set of bus progression offsets can be developed off-line, using the BUS TRANSYT program. These offsets then become the default offsets and have a maximum offset weight associated with them. Thus SCOOT is encouraged (but not forced) to use these offsets rather than those from the offset optimiser.
* Bus Priority using TIRIS (B)
If information on the approach of a bus at a set of signals is available then this information can be used in order to benefit buses, as described below.
In PRIMAVERA, these bus priority components have been integrated with queue management[5] and traffic calming components. The most successful components are as follows:
* Horizontal Queue Model (Q)
This strategy improves the traffic model within the SPOT system so that queue lengths of traffic are estimated accurately. As well as improving the overall performance of the UTC system this also allows strategies such as auto-gating to be implemented as they need an accurate estimate of the amount of space available on each link to store queues.
* Starting & Stopping Wave (W)
Situations in which a queue grows so as to block the entrance to the link at an upstream junction can cause various problems, including loss of junction capacity, increased driver stress and hazards to pedestrians crossing at the junction. Such situations, termed spill-back, are likely to happen in congested networks or on short links. This strategy attempts to control when this spill-back clears from an upstream junction. From work conducted as part of a SERC project[6], it was shown that the least disruptive strategy was to arrange for the starting wave to reach the upstream junction towards the end of either main or minor stage green. For the purposes of PRIMAVERA, the end of minor stage green was chosen as the position in the cycle. If this is effective then there should be a reduction in the number of stops experienced by vehicles thereby reducing the levels of various pollutants and driver stress.
* Speed Advice (S)
The installation of a speed camera on the southern section of Dewsbury road provides a mechanism for enforcing slower progression speeds and producing compact platoons of vehicles that are more easily controlled. This is combined with a VMS sign warning speeding drivers to slow down.
Integration with the UTC systems
Integration with SCOOT. PEEK Traffic modified the SCOOT software to provide priority to detected buses. If information on the approach of a bus at a set of signals is available then this information can be used by the SCOOT optimiser in order to benefit buses. This benefit may take a number of different forms:
(i) Prevent an early termination of the stage which benefits the bus;
(ii) Extend the stage which benefits the bus;
(iii) Recall early the stage which benefits the bus.
Separate weights are available in SCOOT to facilitate each of these actions. These weights are applied at each split optimisation decision (which takes place 5 seconds before a stage change is due) to the merit values associated with the three courses of action. These actions are ADVANCE (bring the stage change forward by 4 seconds), STAY (leave the stage change where it is) or RETARD (put back the stage change by 4 seconds). When a bus crosses the detector, a prediction is made of when the bus will reach the stop line and appropriate weights are adjusted.
It should be noted that there is no absolute guarantee of priority to buses at a junction. At each split decision SCOOT will take into account the merit values associated with decisions on conflicting links in order to reach a decision. This strategy has been designed with the SCOOT incremental philosophy in mind, e.g. the changes due to bus priority are possible outcomes of normal SCOOT operation. The bus detection merely pushes the decision towards one that favours the bus. This is in contrast to a full over-riding priority system that is constrained only by maximum and minimum stage durations.
Integration with SPOT. The SPOT system is based on a decentralised approach. Each intersection aims to minimise a set of cost functions over a rolling horizon of two minutes by communicating with all neighbouring intersections. This process is updated every three seconds. The stage switch over times are only limited by the stage order and the maximum stage durations. This gives a wider range of outcomes than the SCOOT system.
The TIRIS information is used in SPOT to create a prediction of bus arrival at the junction and a new cost is generated in the optimisation process according to whether or not the stages are favourable to the bus. This cost is fed into the total set of costs to be minimised.
Evaluation of the strategies
The strategies were first tested by simulation and the best combination found was implemented in field trials. The NEMIS[1] micro-simulation model was used, being linked up to the appropriate UTC system via an interface developed as part of PRIMAVERA.
Field trial data collection. The surveys at the Leeds site were scheduled to take place with the network operating under different conditions as follows:
(i) Baseline conditions
(ii) SPOT operating alone
(iii) SPOT with the strategies devised for PRIMAVERA
(iv) SCOOT operating alone
(v) SCOOT with the strategies devised for PRIMAVERA
The baseline consists of the network operating under a combination of fixed time and vehicle actuated plans. This is how the network was operating before the PRIMAVERA trials. Six types of survey were carried out between May and December 1994 as follows:
During the trials bus journey times were collected in two different ways. Number plate matching surveys were carried out, with data being collected during the morning peak in the inbound direction. Unfortunately, during these surveys only partial registration plates were collected and it is not easy to distinguish buses with transponders from those without. Therefore this survey has only been used to indicate changes in travel times for the whole bus fleet. Total bus journey times and the delay caused by unscheduled stops were measured by moving observers on buses. Each bus was identified by its full registration plate number. These times were measured for both inbound and outbound journeys over the Dewsbury Road trial site. Data was collected during both the morning and evening peaks. Standard deviations of travel times and delays and statistical confidence levels have been calculated. As it is possible to identify those buses fitted with a transponder from those without, the results for the integrated SPOT and SCOOT strategies only refer to buses fitted with TIRIS transponders. The baseline, SCOOT alone and SPOT alone results refer to any buses, whether fitted with transponders or not. The surveys indicated that about 65% of the buses using the Dewsbury Road were fitted with transponders.
SCOOT Simulation results. For the SCOOT runs the baseline for comparison is the existing road network with signals controlled by the standard SCOOT 2.3 system. Table 1 shows the percentage change in the evaluation impacts over this base case. The changes in travel times of bus services given priority treatment (Bus TT) are also shown. All values refer to the AM peak.
% change P B W+B+S Mean Speed -2.6 -2.1 -3.4 Speeding Time 2.1 2.8 -6.4 Blocking Back 4.6 -7.7 -12.3 Stops 1.0 -0.2 1.7 Delays 1.0 4.6 4.0 Travel Time 2.9 1.9 3.7 Fuel Used 1.4 0.8 0.9 CO Emissions 1.9 2.1 4.3 NOx Emissions 1.2 1.7 3.2 HC Emissions 2.1 2.0 4.1 Bus Travel Time -0.1 2.5 3.7 Bus TT route 2 -1.8 1.4 -4.2 Bus TT route 24 -0.8 2.4 -2.5 Bus TT route 46 -0.3 0.9 -1.9 Table 1: % changes over SCOOTThese simulation results are disappointing. For the SCOOT based strategies in the AM Peak all the strategies perform overall marginally worse than the standard SCOOT system. However, the integrated strategy and the bus progression strategy do manage to reduce the travel times for the priority bus services, without causing much disruption to other traffic.
SCOOT Field Trial results. The results of the number plate matching surveys are shown in Table 2. These surveys cover a two hour period in the morning peak in an inbound direction. The moving observer surveys are shown in Table 3. All the Tables showing field trial results give the number of observations (n), the travel time (s), the standard deviation (sd) and the percentage change in travel time. The integrated SCOOT strategy is referred to as SCOOT +.
The results for the integrated SCOOT strategy appear contradictory. The moving observer surveys indicate a reduction of travel time during the morning peak, while the number plate matching indicates little change. The reason for this is probably that the moving observer studies only considered buses fitted with transponders while the number plate matching considered all buses. Giving priority to buses fitted with transponders at junctions might be producing added delays to buses not fitted, mirroring the simulation results.
Strategy n Travel Time sd % change Baseline 87 406 87 - SCOOT 45 431 113 6.2 SCOOT + 46 407 95 0.2 Table 2: Bus Travel Times - Number Plate Matching Surveys (AM Peak) Strategy n Travel Time sd % change Baseline 35 377 86 - SCOOT 9 379 58 -0.6 SCOOT + 11 348 87 -7.7 Table 3: Bus Travel Times - Moving Observer Surveys (AM Peak) Strategy n Travel Time sd % change Baseline 26 322 98 - SCOOT 9 319 47 -0.9 SCOOT + 12 331 63 2.8 Table 4: Bus Travel Times - Moving Observer Surveys (PM Peak)SCOOT on its own is not very helpful for buses, as on this arterial it imposes strong co-ordination (green waves) for cars between signals. Buses drop out of the co-ordination, often resulting in a red wave for them.
Table 4 shows the moving observer surveys for the PM peak inbound, with buses getting priority in less congested traffic going against the peak flow of outbound traffic. This shows that when travelling against the main flow during the evening peak, bus travel times are not significantly changed when using SCOOT.
The results from both sets of surveys have been combined in Figure 3. Where two figures have been available, the one which has the higher level of statistical confidence has been chosen.
Figure 3: The SCOOT field trial results
% change B Q+B+S Mean Speed 5.6 3.1 Speeding Time 1.8 -8.2 Blocking Back -20.4 6.7 Stops -4.7 -1.6 Delays -8.1 -4.3 Travel Time -5.4 -3.0 Fuel Used -1.2 -1.0 CO Emissions -4.3 -2.7 NOx Emissions -2.5 -1.8 HC Emissions -4.2 -2.4 Bus Travel Time -2.9 -1.7 Bus TT route 2 -6.1 -2.8 Bus TT route 24 -4.5 -1.9 Bus Tt route 46 -3.5 -2.1Table 5: % changes over SPOT
SPOT Simulation results. For the SPOT runs the baseline for comparison is the existing road network with signals controlled by the latest SPOT system. Table 5 shows the percentage change in the evaluation impacts over this base case. Both strategies reduce travel times, environmental impacts and bus travel times by giving priority to buses. Only the integrated strategy has safety benefits by reducing the amount of time spent speeding.
SPOT Field trial results. Again, in the following tables, SPOT+ is the integrated strategy. The results of the number plate matching surveys are shown in Table 6. The moving observer surveys are shown in Table 7.
The integrated SPOT strategy produced a reduction in both bus journey times and journey time variability according to both types of survey. Like SCOOT, SPOT on its own increased bus journey times during the morning peak.
Strategy n Travel Time sd % change Baseline 87 406 87 - SPOT 48 455 118 12.1 SPOT + 52 382 84 -5.9 Table 6: Bus Travel Times - Number Plate Matching Surveys (AM Peak) Strategy n Travel Time sd % change Baseline 35 377 86 - SPOT 38 413 114 9.5 SPOT + 21 348 81 -7.7 Table 7: Bus Travel Times - Moving Observer Surveys (AM Peak) Strategy n Travel Time sd % change Baseline 26 322 98 - SPOT 41 311 73 -3.4 SPOT + 25 290 52 -9.9 Table 8: Bus Travel Times - Moving Observer Surveys (PM Peak)Again the number plate matching survey gives a lower reduction in journey time, reflecting that this sample consists of both equipped and unequipped buses. Results for the PM peak against the flow are shown in Table 8. These show that SPOT can successfully manage to reduce the travel times and journey time variability of these buses. The overall results from both sets of surveys are summarised in Figure 4.
Figure 4: The SPOT field trial results
Conclusions
A cost effective method of giving priority to buses using TIRIS transponders has been developed. This detection system has been shown to be effective when used with either the SCOOT or SPOT UTC systems.
A field trial has shown that during the AM peak, when used with the SCOOT system as part of an integrated strategy on an urban arterial road, that the buses fitted with transponders achieve a reduction in travel time of about 8%. They also have reduced variability in their journey times. Buses not fitted with transponders do not appear to benefit.
When used with the SPOT system as part of an integrated strategy buses fitted with transponders have a similar reduction in journey time and journey time variability. Buses not fitted with transponders also appear to benefit, but not as much as those buses fitted.
REFERENCES
1. FOX, K., MONTGOMERY F.O. AND MAY A.D. (1995) "Integrated ATT strategies for urban arterials: DRIVE II project PRIMAVERA" Traffic Engineering and Control, 36(5), May 1995.
2. BRETHERTON, R.D. (1989) "SCOOT Urban Traffic control System - Philosophy and Evaluation." IFAC Symposium on Control. Communications in Transportation. September 1989, pp 237-239.
3. MAURO, V. AND DI TARANTO, C. (1989) "UTOPIA" CCCT `89 - AFCET Proceedings September 1989 - Paris France.
4. ROBERTSON, D.I. AND VINCENT, R.A. (1975). "Bus priority in a network of fixed-time signals". TRRL Laboratory Report 666.
5. QUINN, D.J. (1992) "A Review of queue management strategies", Traffic Engineering and Control, 33, No. 11.
6. MONTGOMERY, F.O., CLARK, S.D. AND MAY, A.D. (1993). "Use of coordinated signal settings to reduce junction spill-back". IEE Colloquium on Electronics in Managing the Demand for Road Capacity, Digest No:1993/205, pp 10/1-10/10.
PRIMAVERA
home page