当前位置:首页 >> >>


TOPIC AREA: C3 Traffic Control Paper No. 308



Mr. Michael J Smith Mr. Ian W Routledge Dr. Arthur J. Clune

York Network Control Group York Network Control Group York Network Control Group

Page 1

Traffic signals are very common in urban areas and so represent a simple and readily available opportunity to manage an urban traffic network. The project MUSIC, supported by the European Union (DGVII - E2) is designed to show how traffic signals may be used so as to achieve objectives specified by City Authorities. Such objectives may include reductions in congestion for public transport and it is also hoped to show that mode-choice shifts from car to public transport (and from car to other modes) may then be expected in the longer term. Benefits are expected to arise from the use of traffic signal controls designed, taking into account re-routing, in order to create a traffic pattern which increases the efficiency of public transport, while allowing correctly for dis-benefits which may arise from this re-routing. The project involves both modelling and real-life demonstrations in York (UK), Porto (Portugal) and Thessaloniki (Greece). This paper outlines the background to the MUSIC project; the signal control methodology employed and the early modelling and demonstration results achieved.Benefits are expected to arise from the use of traffic signal controls designed, taking into account re-routing, in order to create a traffic pattern which increases the efficiency of public transport, while allowing correctly for disbenefits which may arise from this re-routeing. The project involves both modelling and real-life demonstrations in York (UK), Porto (Portugal) and Thessaloniki (Greece). This paper outlines the background to the MUSIC project; the signal control methodology employed and the early modelling and demonstration results achieved.are expected to arise from the use of traffic signal controls designed, taking into account re-routing, in order to create a traffic pattern which increases the efficiency of public transport, while allowing correctly for disbenefits which may arise from this rerouteing. The project involves both modelling and real-life demonstrations in York (UK), Porto (Portugal) and Thessaloniki (Greece). This paper outlines the background to the MUSIC project; the signal control methodology employed and the early modelling and demonstration results achieved.



Urban Traffic Control (UTC) has become an established part of the traffic engineer’s ‘tool kit’ to improve road conditions in congested cities. Only recently, TRANSYT celebrated its 30th birthday (Robertson, 1997). One of the first computer programmes to deal with traffic control systems, rather than just isolated intersections, TRANSYT has been the starting point for many subsequent developments, and is still one of the standard techniques in use world-wide today. UTC systems have traditionally aimed to reduce congestion, generally expressed as travel times, delays and stops to vehicles, and their control strategies have reflected this. Recent changes in thinking have however shifted the emphasis in urban transport policy from road vehicles to the transport system as a whole, with more attention to other road users (public transport, pedestrians and vulnerable road users (VRU’s)) and minimising the negative impact of traffic (emissions, noise etc.). In addition, it is now widely recognised that continued one-sided improvements to the travel conditions for private vehicles may be counter-productive by attracting even more cars into city centres. A move to placing a greater emphasis on managing transport for the benefit of all travellers was inevitable, however new strategies will be required to support this. Key characteristics of such strategies include:
N explicit attention to all road users, including public transport and VRU’s, with the ability

to benefit specific groups if so required (for example priority for public transport);
N the ability to reflect environmental objectives, related for example to noise and air

quality, over the network as a whole but also distinguishing the needs of particularly sensitive areas and groups;

Page 2

N the flexibility to pursue different, possibly conflicting, objectives in different parts of the

city at the same time or the same part of the city at different times;
N an area wide approach, covering the whole city, rather than just the area controlled by a

UTC system; N a longer term perspective, which allows for, even anticipates, the reactions of road users to control strategies; and N an ability to support and take account of new concepts, such as information systems (pre-trip, in-vehicle or roadside), pricing systems (road pricing, congestion pricing, tolling) and other management systems (for example access control, incident management and parking management). MUSIC (Management of traffic USIng traffic flow Control and other measures), an EC supported RTD project under DG VII’s Fourth Framework Programme, aims not only to provide new tools for developing traffic control strategies to support at least some of these wide ranging policy objectives, but also to provide a mechanism to assess the potential impacts of different policy options on street. MUSIC at current provides tools to help optimise public transport travel times as well as general traffic delay. This paper sets out the principles of managing traffic using flow control and other measures as developed in the MUSIC project and describes its implementation in state-of-the-art micro-simulation based software. Initial simulation results are presented and the planned demonstrations in three cities are described. 2. THE MUSIC APPROACH

MUSIC is a natural progression from mainly academic research carried out over the last 2 decades. Its main objective is to demonstrate that novel methods of traffic control can be used, alone or in combination with other measures, to: N reduce congestion and the environmental impact of urban traffic; N improve the efficiency of urban travel; and N influence modal choice. Control strategies that may be developed and tested within MUSIC, alone or in combination and in roughly increasing order of sophistication, are: N local bus priority measures, which do not take account of route choice decisions in response to the measures; N gating strategies, protecting sensitive areas or defined routes from congestion, but generally ignoring any re-routing effects; N network capacity maximising strategies, encouraging driver route choices which reduce congestion, but without specific regard for road users other than private vehicles; and N novel control strategies designed to interact with other traffic management and control measures, and which explicitly allow for future choices of travellers. These other traffic management measures with which the control strategies could interact include park-and-ride, re-allocation of road space to other road users and road pricing. The development of control strategies in MUSIC has been undertaken as follows: N refinement and testing of appropriate simulation software together with software to translate data from SATURN models into a form usable in the simulation software;

Page 3

N agreement of targets which city authorities would like to achieve (in terms of for

example travel times on corridors or for specific road users, overall levels of congestion, area specific flow levels or improved pedestrian facilities); N simulation of MUSIC effects for three demonstration sites (York, Thessaloniki (Greece) and Porto (Portugal)) using the newly developed software and also more established tools (SATURN); N comparison of the MUSIC results against the city targets and, where necessary, further fine-tuning; and N real-life implementation in the three cities accompanied by appropriate before and after monitoring. The first four steps have now been completed and the paper reports on these. It is hoped to report on the evaluation of the field trials at a later date. 3. 3.1 The MUSIC Concepts General Approach

The use of traffic signals as management tools to reduce congestion and improve traffic and environmental conditions has been advocated for many years; in isolation (Smith, 1980), in combination with information systems (Van Vuren and Van Vliet, 1992) or in combination with road pricing (Ghali and Smith, 1993). Simulation tests reported in these publications, using a variety of network descriptions and simulation models, have generally shown that substantial network benefits can be obtained by an apt choice of control strategy. The concepts behind the MUSIC strategy find their roots in Ghali and Smith (1994). The strategy is implemented in a MUSIC procedure, illustrated in Figure 1. The procedure consists of an iterative process, is described in detail in Appendix A, but the major steps are: N defining city objectives; N translating data from SATURN models into a usable format; N assigning traffic to the network according to current conditions and signal timings and running the MUSIC simulation model; N adjusting ghost prices at the network level and other parameters at the link level within the model to meet the objectives of the city authorities; and N producing the timing plans and testing in the SATURN model. It is assumed that when the timing plans are implemented on street sufficient traffic will reroute in accordance with the predictions to generate the target objectives. The demonstrations in the MUSIC project are designed to check this assumption. 3.2 Ghost pricing

Ghost pricing is road pricing, applied in the simulation model, to affect drivers route choice in that model. For example, within a simulation one can make drivers costs dependent on the time taken to make their journey, resulting in re-routing onto less congested links. In previous network-wide travel time optimisation, ghost prices might have represented the difference between actual and marginal link costs. With more diverse objectives, some

Page 4

incompatible and distributed around the network, ‘good’ values of ghost prices are harder to find and their interpretation is less straightforward. Ghost pricing strongly encourages beneficial traffic patterns in the simulation model. In the practical implementation other means, particularly traffic signals, (but in extremis also other, physical measures such as traffic calming or bus gates) must be employed to encourage as many road users as possible to choose and keep to their new routes. Hence, the control strategy employed is crucial in maintaining the potential benefits when the ghost prices are removed. Under the P0 signal setting strategy, each stage is given a "pressure", P=s1d1+s2d2+ … where s is the saturation flow for a stage and d is the average delay. Signals are then altered so as to give more green time to stages with higher P, until the "pressures" have equalised. It has been proved in a simple model (Ghali and Smith, 1993) that under certain delay assumptions the combination of delay based pricing and the P0 signal control strategy developed at York University (Smith, 1980) leads to efficient utilisation of network capacity. This result is used in the current MUSIC version to help determine both the initial ghost-prices and the associated signal adjustments. Hence, the current MUSIC version employs system-wide delay-weighted ghost-pricing and applies the P0 control strategy at all signalised junctions. Initially a single weight is applied to all delays experienced at signalised junctions, expressed as pence per minute of delay. For each price level to be tested, traffic is loaded onto the network and allowed to find routes to destinations. The traffic signals are then adjusted using P0. This process is continued for several iterations, each representing a day’s worth of traffic for the time period considered (usually the morning peak). After a fixed number of iterations, the signals are frozen, the prices are discarded, and the model iterated until it reaches equilibrium. Travel times, route flows and delays are then measured. The MUSIC procedure searches through the range of reasonable values for this weight until an optimum value for the specific combination of city objectives has been found. The traffic signal timings which arise have been called proactive timing plans. 4. NEED FOR APPROPRIATE STRATEGIES

Current transport policies endeavour to widen the range of issues taken into account when developing policy options (environmental as well as traffic conditions, vulnerable road users and public transport as well as private vehicles, long term sustainability rather than short term improvements). We suggest that strategies that actively aim to influence travel patterns so as to support transport policy have a pivotal role to play. P0 is a prime example of such a strategy. Many current UTC systems apply variations of delay minimising strategies, optimising the green times to fit current flow patterns, or those predicted to occur within the near future. Developments in short-term forecasting have enabled these strategies to be adaptive, highly dynamic and indeed quite sophisticated. However, these are essentially reactive strategies, accepting and responding to observed driver choices, sometimes inadvertently encouraging poor route choice behaviour, with little attention paid to wider effects on the urban

Page 5

transport system. We would therefore suggest that responsive UTC systems may not always be the best way to support evolving policy and that at certain times, for certain parts of the network, MUSIC developed pro-active fixed time plans offer a better approach. The control strategy represented by MUSIC is aims to generate beneficial flow patterns (with respect to those objectives identified by the City Authorities). At the moment, this strategy is based on three main components: delay-based ghost-pricing, P0-based signal timing adjustments and final manual fine-tuning. The signal control strategy employed seeks to support or enhance the changes in travel patterns induced by the pricing regime. In due course these essentially heuristic components should be replaced by actual optimisers, and relevant research is in progress. Two further considerations apply: N MUSIC signal plans are in principle non-responsive: guiding traffic, rather than reacting to traffic. However, in practical implementations, minor green time adjustments to random flow fluctuations could in some situations be allowed for; and N because the P0 strategy only accounts for road-based traffic, further manual refinements are required after the iterative (ghost pricing/signal adjustment/re-assignment) procedure, to fine-tune the resulting signal timings in the context of more detailed overall, possibly multi-modal objectives. This is discussed in the next section. 5. SIMULATION APPROACH

SATURN network models have been set up for the three demonstration sites: York, Thessaloniki and Porto. They have been designed, calibrated and validated to the standards required in the Urban Traffic Advice Manual (DOT, 1996). All MUSIC signal timing designs and analyses have been carried out using the newly developed STEER (Signals/Traffic Emulation with Event-based Responsiveness) microsimulation software (see for example Ghali, Smith and Clegg, 1995). A translation program was been developed to translate data from the SATURN models into a format usable in STEER. As the STEER data structure is similar to that of CONTRAM, this translation package may also be used for direct SATURN - CONTRAM translations. A simulation approach was adopted that would make best use of the strengths of the two models: SATURN N well-established and respected in international practice N substantial existing experience within participating organisations N extensive analysis facilities N existing SATURN models in many towns and cities are available. STEER N detailed representation of traffic dynamics and individual vehicle behaviour N representation of UTC, pricing and signal control strategies N suitable for continuous development within the project

Page 6

The network model is created and calibrated in SATURN. The optimisation is carried out using the MUSIC procedure, implemented in the STEER software. The calculated signal timings are analysed within STEER and then fed back into SATURN, allowing its extensive analysis facilities to be used. The MUSIC procedure consists of three possible iterative loops, combining automated procedures with an ability to use local knowledge and engineering judgement to refine the signal timings in the light of the stated targets: (a) the loop within the MUSIC procedure itself, iterating between ghost prices, signal adjustments according to the P0 control strategy and re-assignment (as shown in Figure 1 and discussed in section 3); (b) a second loop, in which parameters are fine-tuned to achieve a better fit with city objectives (also shown in Figure 1); and (c) a final possible loop that may use the results of the feedback into SATURN to fine-tune the signal plans even further. The basics of the MUSIC procedure have been discussed in Section 3. The fine-tuning of parameters in loops (b) and (c) may concern, for example, the following: N increase or decrease of cycle times to affect capacity, but also waiting times for crossing pedestrians or the effectiveness of bus priorities; N differential weights for delay-based ghost prices in different areas, dependent on the sensitivity of areas and how far the city objectives have been achieved there; N manual adjustments to green splits where the automated procedure has failed to calculate reasonable settings; N adjustments to offsets to achieve better progression where desired; N increase in lost time per cycle, re-allocating more green time to pedestrians. The availability of equivalent network models in alternative software suites enables a rare check on the simulation results, particularly related to the influence of assumptions. In a similar context Van Vuren and Van Vliet (1992) remarked how, for example, assumptions about delay functions affect not only the overall predicted effectiveness of alternative control strategies but (in the MUSIC context even more relevantly) to completely different ‘optimum’ signal timings. For implementation on street this consideration is critical. The use of several different models has proved considerable added confidence in the possible benefits of the proposed signal timings. 6. SIMULATION RESULTS

For all three demonstration sites, objectives have been obtained from the city authorities, and translated into quantifiable targets. As can be seen from Tables 1, 2 and 3 these range from improved park and ride bus journey times in York, to traffic flow reductions in pedestrian areas in Porto, to reductions in delays on strategic routes in Thessaloniki. The MUSIC procedure has been applied in all three networks to determine signal timings that deliver these (potentially conflicting) targets. In none of the cases have the full iterative loops (a) (b) and (c) been applied. For York loops (a) and (b) have been run. Where loop (a) aims at the optimisation of the city-wide network, the current emphasis for implementation is on localised improvements to the operation of the Park and Ride service. Hence, the presented results are based on the application of loop (b) only. For Thessaloniki and Porto no further fine-tuning of the automated signal plans has taken place as yet. Hence
Page 7

STEP 1 Translate SATURN models City SATURN model

STEP 2 Agree detail city requirements

STEP 3 Set parameters and price

STEP 4 Run STEER simulation model

STEP 6 Adjust price

STEP 5 Price range adequate ?


STEP 8 Adjust parameters


STEP 7 City requirements achieved? NO YES STEP 9 Test timing plan in city SATURN model

STEP 10 Summarise results

Page 8

these results are based only on the results of loop (a). It will be seen that even then many of the city objectives can be met. 6.1 York

Hull Road is an important East West corridor into York city centre; one of York’s Park and Ride sites is situated here, just off the A64 Ring Road as shown in Figure 2. Currently, all traffic into the City Centre travelling along Hull Road experiences little or no delay until the two lane carriageway reduces to a single lane at the approach to the existing Melrosegate traffic signals. Here, significant delays occur in the AM peak. The objectives of the City for travel conditions along this corridor are set out in full in Table 1 but can be summarised as follows: N reduce delays for park and ride buses, without significantly increasing delays for other westbound traffic on Hull Road; N reduce delays for all vehicles on Hull Road at the Melrosegate signals travelling into York; N avoid excessive diversions through the village of Murton. These objectives are not necessarily compatible. For example decreasing green time to general traffic can help bus journey times but lead to much higher diversion flows though sensitive areas. Loop (b) in the MUSIC implementation has been used to investigate a wide range of control strategies and associated timings, consisting of a combination of gating and bus priority. Table 1 presents a summary of simulation results as assessed by SATURN. It seems that, according to our modelling, the bus-gate scheme as envisaged at present will not have a beneficial effect on bus travel times, with large dis-benefits for other groups. It seems likely that the +50% figure for delay at Melrosegate is an artefact of the SATURN model since in both CONTRAM and STEER the delay at Melrosegate was very little changed (+1-2%) from the base case. Also the signal timings at Melrosegate were altered (by York City Council) between the base case and after models so this result is not directly comparable. It may be that the change to the signal timings at Melrosegate is swamping any benefit from the bus gate scheme. These results are however very poor in terms in the objectives set, especially given the lack of benefit to bus journey times, and the implementation of the scheme may need to be considerably altered in order to gain the benefits planned.
TABLE 1 SUMMARY OF YORK REQUIREMENTS AND RESULTS Description Measurable Target Hull Road from the Park and Ride site exit to just west of the Melrosegate traffic signals Melrosegate straight ahead movement Hull Road from the Ring Road to just west of the Melrosegate traffic signals Diversion routes: Murton, Tang Hall and University Road M1 M2 M3 Bus journey time All vehicle delays Car journey time Traffic flow Reduction of 20% Reduction of 50% Limit increase to 10% Limit increase to 30% Achieved +3% (+50%) +25%



Page 9


Tang Hall Osbaldwick Lane Tranby Avenue Hull Road (A1079) B Bus Gate Traffic Signals Field Lane C Park and Ride Site Traffic Signals Park and Ride Ring Road (A64) Murton

Tang Hall Lane A Melrosegate Traffic Signals

University Road


Page 10



Porto is the second largest city in Portugal. The central core has a dense traffic system without any clear hierarchy. As in York, the limited river crossings present bottlenecks. Both and Inner and Outer Ring Road are present. The objectives agreed with the city authorities seek, as can be seen in Table 2, a balance between the following: N reduction in vehicle journey times on major roads; N reduction in traffic flows in pedestrian areas, providing both a more pleasant and safer environment, and better crossing opportunities through a re-allocation of green time to pedestrian phases; and N reduction in bus travel times. The automated MUSIC procedure has been applied without any further fine-tuning. Table 2 shows the simulation results, which have been obtained with very low optimum ghost prices of 1.5 pence per minute. Again, these are presented separately for the STEER optimisation and the feedback of calculated timings into SATURN, whilst also assessing the attainment of the specified city objectives (based on the STEER results). The following observations can be made:
N the estimated network wide benefits in delays and travel times, at 5-8%.are much lower N


than for Thessaloniki; feedback of the timings into SATURN reduces these estimates of overall benefits to 12%, substantially lower but still a positive result supporting the implementation of the calculated signal plans in reality; nearly half the city’s objectives are met completely, even though no further adjustments to the timings have been made (particularly where these may be conflicting); in one case the effects of the signal plans actually exceed the authorities’ objectives, which may be remedied by manual adjustments (if so desired); in a further three cases the objectives are nearly achieved (correct direction but insufficient absolute change); fine-tuning may make these attainable; and six cases, some 40%, require further attention, as the resulting changes in measurables are opposite to the desired effects.

Page 11


  !  ! ##   ! ! ##

Map 1.

The Porto demonstration site

Page 12

TABLE 2: SUMMARY OF PORTO REQUIREMENTS AND RESULTS Network Wide Results (existing timings as base) Optimal MUSIC timings in Optimal MUSIC MUSIC model timings in SATURN Total travel time -5% -1% Total delay -8% -2% Total free flow -1% -1% Total travel distance 0% 0% Attainment of City Objectives Measurable Target Achieved using MUSIC M1 Vehicle journey time Reduction, at least 10% ++ M2 Traffic flow Reduction, not more than xx 10% M3 Vehicle journey time Reduction, at least 10% + M4 Traffic flow Reduction, not more than ++ 10% M5 Vehicle journey time Reduction, at least 10% x M6 Traffic flow Reduction, not more than x 10% M7 Vehicle journey time No increase ++ M8 Vehicle journey time No increase xx M9 Bus travel time Reduction, at least 10% ++ M10 Bus travel time Reduction, at least 10% x M11 Bus travel time No increase x M12 Bus travel time No increase ++ M13 Bus travel time No increase x M14 Pedestrian green time as Increase, at least 5% + proportion of cycle time M15 traffic flow Reduction, at least 10% xx M16 traffic flow Reduction, at least 10% ++ M17 traffic flow Reduction, at least 10% x ++ = well met, + = met, x = not quite met, xx = not met at all



Thessaloniki is Greece’s second city. Traffic to and from the central area uses a number of high capacity routes, which at peak times become very congested. The city objectives are concentrated on these routes where both throughput and delay are to be improved. The modelled network is extensive, some 500 nodes, half of which are signal-controlled. Implementation of the MUSIC procedure established pro-active signal plans for these junctions; no further manual fine-tuning has yet been attempted. The optimum delay weight (or ghost price) was zero, so that in effect a straightforward P0 application was carried out. A partial reason for the price of zero was that in the original SATURN model, traffic costs where based only on time and not on time and distance. Since adding to the time costs can not alter the ratio between distance and time costs the MUSIC method of pricing delay is less effective here.

Page 13

Thessaloniki Music : Refined City Objectives


6 7 10

11 8 9
Th ess alon iki Net w ork Pou te 1 Rou t e2 Rou t e3 Rou t e4 Rou t e5 Rou t e6 Rou t e7 Rou t e8 Rou t e9 Rou t e10 Rou t e11



3 2

Map 2.

Thessaloniki demonstration site

Page 14

Table 3 summarises the impact of these timings on travel conditions in Thessaloniki, separately for the STEER optimisation and the feedback of calculated timings into SATURN, and also assessing the attainment of city objectives (based on the STEER results). The following observations can be made: N the calculated possible benefits in terms of total delays and travel times in the network are significant: 22% and 17% respectively. Feedback of timings into SATURN approximately halves these estimated benefits, but still retains reductions in delays and travel times in the order of 10%. N although no subsequent adjustments to the timings have been made (to attain better the specific city objectives - particularly where these may be conflicting), for two-thirds of the measurables the targets have been met. N in a further 10% of measurables the objectives are nearly met (correct direction but insufficient absolute change); fine-tuning may make these attainable. Whilst less than a quarter of the measurables require detailed attention (where the resulting changes in travel time and or flow are opposite to the desired effect). However, it must be stated that the existing timings in Thessaloniki are probably far from optimal, and so the large benefits gained here might not be generally achieved.
TABLE 3 : SUMMARY OF THESSALONIKI REQUIREMENTS AND RESULTS Network Wide Results (existing timings as base) Optimal MUSIC timings in Optimal MUSIC MUSIC model timings in SATURN Total travel time -17% -9% Total delay -22% -10% Total free flow 0% -2% Total travel distance +2% +2% Attainment of City Objectives Measurable Target Achieved using MUSIC M1 Journey time Reduce by 10 - 20% ++ M2 Flow Increase up to 5% x M3 Journey time Reduce by 10% x M4 Journey time Reduce by 10% ++ M5 Flow Increase up to 10% + M6 Journey time Reduce by 5% x M7 Flow Increase 5 - 10% + M8 Journey time Reduce by 10% ++ M9 Journey time Reduce by 10 - 20% xx M10 Journey time Reduce by 15-20% + M11 Journey time Reduce by 10 - 20% ++ M12 Journey time Reduce by 10 - 20% ++ M13 Journey time Reduce by 10 - 20% xx M14 Journey time Reduce by 10 - 20% ++ M15 Delay Reduction of 2 minutes ++ M16 Delay Reduction of 2 minutes ++ M17 Delay Reduction of 3 minutes x M18 Delay Reduction of 2 minutes ++ ++ = well met, + = met, x = not quite met, xx = not met at all

Page 15



As part of the evaluation of MUSIC the on-site implementation will be carefully monitored. Travel times and other objectives as given in Tables 1,2 and 3 will be measured on site both before and after the implementation of the MUSIC signal plans. The results from these studies should provide valuable data on the performance of the MUSIC project. Before surveys have been undertaken in York, the bus gate implemented on street and the timings for the bus gate are currently being set. The after study should be completed by March 1998. Before studies are underway in Porto and Thessaloniki and should be completed by March 1998. 8. SUMMARY

Simulations have been carried out in the three demonstration sites, and are currently being refined. Despite the three cities having widely differing objectives, the MUSIC strategy has achieved benefits in all cases, overall, in the specified sub-areas, for many of the explicit targets and for the separate modes identified. The feedback process of calculated signal timings into an established assignment package based on rather different assumptions than the STEER software supports the robustness of the estimated signal timings for actual onstreet implementation. All modelling results refer to the “base case” and we propose to consider other comparators. The three cities vary in size, and also in the authorities’ targets. In York, small scale improvements to the P&R operations are sought, but the MUSIC simulation results show that achieving benefits will require further fine-tuning of the scheme to be implemented. Recent work by Hounsell and Wu (Hounsell and Wu 1998) confirms that some types of bus gate scheme do indeed produce a dis-benefit to the busses. In Porto, a balanced improvement is sought between conditions for car drivers, public transport passengers and pedestrians. Again, the MUSIC results indicate that many of these can be achieved by application of a control strategy that (in its initial operation) concentrates on longer term capacity improvements for road traffic, generally through re-routing to longer, but faster routes. The resulting reduction in congestion in central areas benefits the fixed-route public transport, whilst also more green time can then be made available to pedestrians (with an additional benefit of reduced vehicle flows in sensitive areas). Opportunities for further, manual fine-tuning have been identified. In Thessaloniki, all benefits sought by the authorities accrue to road traffic. Many of these are achieved in an initial MUSIC application, with some scope for further manual fine-tuning. Bus routes along the improved corridors share these benefits. Concluding, the MUSIC procedure can be seen to be flexible and powerful. The automated ghost pricing - signal adjustment - re-assignment procedure may be employed to produce initial “cartoon” signal plans. Using these as a starting point, subsequent manual refinements enable the traffic engineer to take a city-wide whilst still retaining influence over local impacts in the light of often conflicting policy objectives.

Page 16

The need to take a system-wide view when optimising traffic control (particularly allowing for road users’ reactions to the signal timings) has been advocated for over 20 years (for example Allsop, 1974 and Maher and Akcelik, 1975). This has become particularly significant in the current drive for UTMC system development and implementation (see Routledge et. al. 1996). The basic foundations of the MUSIC UTMC strategy were laid around the same time (Smith, 1980). Despite numerous successful simulations under different model assumptions this is the first time that this kind of pro-active signal timings will be implemented on-street. We hope to report on the results of this implementation in a future paper. 9 ACKNOWLEDGEMENTS

MUSIC is funded under the Fourth Framework Programme by Directorate General VII(E2) of the European Commission. MUSIC consortia members are the University of York, Hague Consulting Group, the City of York Council, TRIAS SA (Greece), and the Universities of Porto and Coimbra in Portugal. We acknowledge the contribution of Richard Clegg in the development and operation of MUSIC in the STEER software package. 10 REFERENCES

Allsop, RE (1974) “Some possibilities for using traffic control to influence trip distribution and route choice” in Buckley, DJ (ed): Proceedings of the 6th International Symposium on Transportation and Traffic Theory, Sydney, AH and AW Reed Pty, Artarmon, NSW, pp 345-372 DOT (1996) “Traffic appraisal in urban area”, DMRB v12.2.1, HMSO, 1996 Ghali, MO and Smith MJ (1993) “Traffic assignment, traffic control and road pricing”. Proceeding of the 12th International Symposium on the Theory of Traffic Flow and Transportation (Ed. CF Daganzo). Elsevier, pp147-170. Ghali, MO and MJ Smith (1994) “Designing time-of-day signal plans which reduce urban traffic congestion” Traffic Engineering and Control, Vol. 35, No 12, pp 672-676 Ghali, MO, MJ Smith and RG Clegg (1995) “A new dynamic micro-simulation/assignment model and new estimates of traffic control system performance including bus priority effects” PTRC 23rd European Transport Forum, Warwick, Proceedings of Seminar E (Transportation Planning Methods), pp 153-163 Hounsell, N and Wu, J (1998). “Models for pre-implementation evaluation of bus priority using pre-signals”. Pre-print, presented at U.T.S.G. 1998. Maher, MJ and R Akcelik (1975) “The redistribution effects of an area traffic control policy”, Traffic Engineering and Control, Vol. 16, No 9, pp 383-385 Robertson, DI (1997) “The TRANSYT method of co-ordinating traffic signals” Traffic Engineering and Control, Vol. 38, No 2, pp 76-77 Routledge, I, S Kemp and B Radia (1996) “UTMC: The way forward for urban traffic control” Traffic Engineering and Control, Vol. 37, No 11, pp 618-623 Smith, MJ (1980) “A local control policy which automatically maximises the overall travel capacity of an urban network”, Traffic Engineering and Control, Vol. 21, pp 298-302 Smith xxxx Van Vuren, T and D Van Vliet (1992) “Route choice and signal control”, Avebury, Aldershot, 1992

Page 17


Step 1: the Translation Package was used to automatically translate the physical SATURN data into a form suitable for use by the simulation models. Step 2: the Evaluation Framework describes the measures to be used to assess the benefits of the MUSIC software. In order to develop appropriate proactive timing plans these general measures were expanded into detailed and quantified requirements for each city. Step 3: the existing traffic signal timings and flows/queues were reviewed in light of the city detail requirements and initial parameters set, in particular the cycle times, offsets, green splits and intergreens. Usually the estimated existing timings were utilised as the initial stating point. The ghost price was also initially set to zero. Step 4: the MUSIC microsimulation model was used to produce results. The outputs produced for each city were dependent upon the detail city requirements but could include: N total vehicle delay in the network; N total travel time in the network; N travel time on specific links N traffic flows on specific links; N queuing delays; and N pedestrian benefits. Step 5: the various outputs produced were then plotted against ghost price and visually inspected to ensure that the price range was adequate and included the optimum price with sufficient results either side of the optimum to give confidence that the true optimum had not been overlooked. Additional prices around the mean were also produced. Step 6: when determined necessary to adjust the ghost prices this was done at a fairly course level until the necessary range of prices was achieved. Infill prices were based on the previous results but would typically halve the price interval until the desired density was obtained. Step 7: Having determined results for an adequate range of ghost prices, the results were then inspected and compared against the detail city requirements, bearing in mind the full range of requirements, some of which were incompatible. If the results were found not to meet the city requirements the parameters were adjusted (Step 8) and the model rerun (Step 4). If however the results were considered the best that could be achieved then Step 9 was implemented. Step 8: if Step 7 determined that the results did not meet the city requirements then the parameters were adjusted to produce better results. This could include: N adjusting the cycle times at specific traffic signal installations; N adjustment of timings at individual traffic signal installations; and N applying different ghost prices to different areas or traffic signal installations.

Page 18

After adjustment of the parameters Step 4 was implemented. Step 9: the optimum timing plans produced by the microsimulation model were then fed back into the city SATURN models and results obtained for the range of requirements identified in Step 4. This was done to give confidence to the cities that the results produced would give the anticipated benefits. It was never envisaged that the SATURN results produced using the microsimulation model optimised timings would mirror the results from the microsimulation model but it was considered that they could confirm the general trend. Step 10: a summary of results was then prepared comparing network wide results including: N SATURN model outputs for estimated existing traffic signal timings; N MUSIC timing plan results from the timings produced by the microsimulation model; and N SATURN model outputs for traffic signal timings produced using microsimulation model.

Page 19

Traffic flow model 交通组成 traffic diversion (车辆)超载 弹性系数 交通流模型...of service Traffic control and managementsite visit site visit 服务水平 交通...
Traffic Accidents.doc
Besides, the weak supervision and management of traffic control sectors is a...problem, it is high time for us to take more effective measures to solve...
The press, publishing, radio, television and other relevant units, have ...?Article 7 of the road traffic safety management work, should strengthen ...
4.Air Traffic Flow & Capacity Management_图文.ppt
Management Strategy Air traffic Flow and ...available capacity of the air traffic control (ATC...? Such measures frequently result in delays of ...
So, too many vehicles, too old or narrow roads, and any other immature ...The last cause is the insufficient management of the traffic systems and ...
How to Solve Traffic Problems in Big Cities英语作文.doc
congestion, we need to control the traffic flow....other solution: the different type of the cars'...Whether in the traffic management part or the ...
traffic jams.doc
So, not only should the government make some measures, also citizens should...And the government also should make effective management of road traffic. To...
...the differences in traffic accident between China and U.S.....doc
and other factors,there are many bicycles are on...traffic management ideas and measures can not keep...traffic flow models, mixed person and vehicle and...
空中交通管理 ATM Traffic Management_图文.ppt
Congestion fluctuation in traffic flow and faults within the network Traffic Control Traffic Management Congestion Control Traffic Management forms: Signaling ...
How to Solve the Problem of Heavy Traffic.doc
This can decrease the traffic flow. But on the other haad, this will ...So people begin to thrash out the way of traffic crowding and jamming, ...
how to avoid traffic accidents_图文.ppt
The windshield of George is very dirty, what should he do to avoid ...Watch the traffic lights and the other traffic carefully at crossings when ...
Automatic Traffic Density Estimation and Vehicle Cl....pdf
for signal control and effective traffic management...Several other vehicle detectors such as loop, ...has been developed using traffic magnetic sensors....
Traffic Management_图文.ppt
(PCR>SCR) Nrt-VBR UBR Other traffic PCR SCR ...Management Congestion Indication, Flow Control Round...and control the amount of incoming ATM traffic. ...
How to Solve the Traffic Problem1.doc
At the same time, legislative measures should be introduced and pedestrians ...Such as building more roads and speeding up the flow of traffic. We also...
...Integration of Adaptive Traffic Control and Trav....pdf
Traffic Management, Real-time Control, Optimization...portion of the horizon by using traffic detectors...revised dynamically when more recent real-time ...
How to salve the traffic problem.doc
The other one , traffic control and law enforcement need to be improved ....A dozen of measures are supposd to take to prevent traffic jam from ...
M7 unit 4 Project Traffic Accidents and Road Safety....doc
B.Heavy traffic. C.Overtaking other vehicles. D.Not paying attention to surrounding traffic. 6.What does “This lack of control is an invitation to ...
...is the best way to solve growing traffic and pol....doc
of petrol is the best way to solve growing traffic and pollution problems....What other measures do you think might be affective? With the rapid ...
Fundamentals of Traffic Flow.pdf
Fundamentals of Traffic Flow From single vehicle data...c optimization measures like on-ramp control, ...c models. On the other hand, some hardly ...
Beijing traffic.doc
using the northern section of the 2nd ring road...The city’s poor traffic management and physical ...continuous traffic flow, and smooth communication betwe...

All rights reserved Powered by 甜梦文库 9512.net

copyright ©right 2010-2021。