The 13th International Conference “RELIABILITY and STATISTICS in TRANSPORTATION and COMMUNICATION - 2013”

Proceedings of the 13th International Conference “Reliability and Statistics in Transportation and Communication”
(RelStat’13), 16–19 October 2013, Riga, Latvia, p. 117–127. ISBN 978-9984-818-58-0
Transport and Telecommunication Institute, Lomonosova 1, LV-1019, Riga, Latvia

VARIABLE SPEED LIMITS.

REVIEW AND DEVELOPMENT OF AN AGGREGATE
INDICATOR BASED ON FLOATING CAR DATA

Alvaro Garcia-Castro, Andres Monzon

Transportation Research Centre (TRANSyT), Universidad Politecnica de Madrid
Escuela de Caminos, Canales y Puertos, Avda. Profesor Aranguren s/n. Madrid. Spain
Phone: +34 91 336 52 34; fax: +34 91 336 53 62; e-mail: [email protected]

Changing factors (mainly traffic intensity and weather conditions) affecting road conditions require a suitable optimal
speed at any time. To solve this problem, variable speed limit systems (VSL) – as opposed to fixed limits – have been developed in
recent decades. This term has included a number of speed management systems, most notably dynamic speed limits (DSL).
In order to avoid the indiscriminate use of both terms in the literature, this paper proposes a simple classification and offers
a review of some experiences, how their effects are evaluated and their results
This study also presents a key indicator, which measures the speed homogeneity and a methodology to obtain the data
based on floating cars and GPS technology applying it to a case study on a section of the M30 urban motorway in Madrid (Spain).

Keywords: effectiveness indicator, variable speed limits, dynamic speed limits, GPS application, speed management, floating car
data

1. Introduction

The main tool used by authorities to manage speed is the setting of speed limits, which tend to be
fixed. However, the optimal speed cannot remain constant at all times, as the road conditions are affected
by numerous factors, mainly traffic intensity and weather conditions [1].
Speed can be regarded as a key factor that directly affects certain aspects of the road such as traffic
performance, road safety and environmental externalities.
1) Traffic performance
Together with intensity and density, speed is one of the key factors determining road capacity. At a
critical speed and the corresponding critical intensity or density, the state of flow will change from stable
to unstable and, speed differences and braking process can therefore lead to congestion and reduced road
capacity [2].
2) Road safety
It is generally accepted that high speeds involve a high risk to road safety. This idea is supported
by a large number of studies which highlight the relationship between speed and road safety. For instance,
ref. [3] shows an extensive review of 98 studies containing 460 estimates of the relationship between
changes in speed and changes in the number of accidents or accident victims, concluding “the relationship
between speed and road safety is causal, not just statistical”.
3) Environmental externalities
Apart from vehicle technologies, speed is a very important factor determining negative
environmental effects such as CO2 emissions, pollutants [4] and noise [5].

The concern of traffic authorities to adapt traffic speed to changing road conditions has led in
recent decades to the development of variable speed limits (VSL).

2. Variable Speed Limits. Classification

VSL is a broad term that includes many speed management systems with different motivations and
control algorithms. VSL can be defined simply as speed limit management systems, which are time
dependant. Some authors confine the term VSL to systems, which utilize traffic detectors to determine the
appropriate speed [6]; however, this fails to take into account the existence of VSL that operate following
prefixed calendars or timetables based on historical data. It is thus necessary to classify VSL as follows:

117
Session 3. Transport Systems

2.1 Scheduled Variable Speed Limits (SVSL)

These are VSL that depend on a pre-established calendar or timetable. Among these, the following
types can be identified:

Seasonal Variable Speed Limits

These are applied to a specific type of road and set the speed limit during a particular season, with
the most common being the winter/summer speed limits.
An example can be found in the Nordic countries due to their extreme weather conditions during
the winter months. In Finland the reason for lowering winter speed limits is primarily the adverse road
and driving conditions [7].
Experiments involving the setting of seasonal speed limits for safety reasons can be also found in
the northern states of the U.S.A. For instance, the Wyoming Department of Transport first implemented
the seasonal speed limit for six months beginning on October 15, 2008 [8].
In the Austrian region of Tyrol during the winter of 2006/2007, the speed limit was temporarily
reduced on the Inn Valley Motorway from 130 km/h to 100 km/h, mainly due to high levels of air
pollution during previous winter seasons [9].

Hourly Variable Speed Limits

These are mainly applied to prevent or reduce certain negative externalities in a specific road
section or street at particular times.
Experiments of this type can be found in some German or American cities where authorities have
implemented VSL in school areas in order to reduce speed when schools are open or at exactly the times
children are arriving or leaving [10], [11].
There are also experiments related to noise reduction during night hours in residential areas or
close to hospitals and other facilities. In the city of Berlin [10], speed is limited during the night hours to
30 km/h in residential or mixed areas.

Figure 1. Proposed classification of Variable Speed Limits

2.2 Dynamic Speed Limits (DSL)

The term “dynamic” implies a force which produces a change in state or movement. In this case,
the “forces” that produce changes in speed limits are the conditions in and around the road. Therefore,
Dynamic Speed Limits can be defined as a type of Intelligent Transport System (ITS) which produces
changes in speed limits in response to accurate information regarding road, driving, weather and/or
environmental conditions [2].

118
 The 13th International Conference “RELIABILITY and STATISTICS in TRANSPORTATION and COMMUNICATION - 2013”

In practice, the system consists of dynamic message signs (DMS) deployed along a roadway and
connected via a communication system to a traffic management centre [12]. After data processing and
speed limit calculation, the new speed limit information is displayed on these DMS.
Pure manual control methods are based simply on a protocol that the operators activate when one
or more levels (traffic intensity, visibility, air pollution, etc.) exceed the pre-set thresholds.
The concept of automatic DSL is based on various approaches, ranging from basic protocols
according to particular thresholds (similar to manual methods) to complex algorithms based on multi-
objective optimization, game theory, predictive control and genetic algorithms [13]-[16].

3. Evaluation of Variable Speed Limits

DSL are being implemented worldwide; however their effects are not yet clearly defined, and in
some cases their benefit is not fully proven.
Based on international experiments and research studies, we summarize the way in which DSL
affect the parameters of traffic performance, road safety and environmental levels, and the variables that
are used to assess their effectiveness.

3.1 Traffic Performance

With regard to traffic performance and traffic flow behaviour, there are several parameters which
may be affected by the implementation of DSL, including particularly speed and capacity.
The reduction in average speed and speed variations depends largely on the type of speed limits
imposed (mandatory or recommended) and their enforcement. Most DSL operate as mandatory limits,
such as the M25 controlled motorway around London [17], although there are also some systems with
recommended speed limits, such as the Motorway Control System (MCS) on the E4 in Stockholm [18].
These systems are based on the capacity increase that occurs when speed and speed variations are
reduced by high flow levels. Moreover, speed homogenization reduces the number of acceleration and
deceleration manoeuvres and therefore the oscillations in traffic flow [19]. Ref. [20] shows that under
certain congestion conditions, speed determines density; based on this observation, the relationship
between density and speed can be estimated depending on speed limits.
The reduction of speed limits has a considerable effect on the speed differential between lanes. In
[2], the conclusions show that it can be said that dynamic speed limit systems do increase the
homogeneity of the driving speed.
Based on computer simulators, some authors evaluate positively the effects of DSL on traffic
performance. Reference [21] shows the simulation of a number of types of DSL scenarios, and the results
indicate that the benefits of DSL are obvious when the traffic volume is equal to or greater than
2,800 veh. in a double-lane freeway. Ref. [13] simulates the effects of DSL on the prevention of
congestion caused by shockwaves, obtaining a reduction in total travel time of 21.7%.
Germany has a long tradition of implementing VSL, and in particular DSL. The first experiment
was implemented in 1965 on the A8 motorway between Munich and Salzburg, with good results in terms
of harmonization and reduction of speed differences between lanes. These results and many others from
German motorways can be found in [22]. Among these cases, we can highlight the report on the A5
motorway in Frankfurt. Based on data from video recording and induction loops, the authors found a
significant increase in the empirical maximum traffic intensity in the southbound direction from
5,200 veh/h. to 5,900 veh/h. (about a 10% increase). However other studies in the Netherlands [23]
estimate the capacity increase at around 2%.
The M25 in the U.K. can also be highlighted as a successful implementation of DSL. During the
first year of operation, a section of this controlled motorway absorbed a 1.5% increase in throughput over
5-hour peak periods, without any detectable increase in congestion levels. Traffic conditions have
improved as a result of the reduction in frequency and severity of shockwaves. The study [17] revealed a
reduction of over 25% in the typical number of shockwaves during the morning peak period. It has also
been observed that the traffic is now more evenly spread across all four lanes.
Reference [18] studied the application of DSL on the E4 motorway through Stockholm, revealing
that lane changes were reduced by over 50%, and that lane distribution became more balanced after the
implementation. However, this phenomenon can have negative effects in sections with a high density of
on-ramps, as this will lead to smaller gaps in the traffic on the outside lane, making the merging process
more difficult and therefore creating congestion on the on-ramp [24].

119
Session 3. Transport Systems

Experiments were conducted on the ASF (Autoroute du Sud de France) in France in the summer of
2004, with the implementation of an innovative traffic control system on the A7 motorway, which
includes DSL. In the southbound corridor, the use of progressively slower speed limits depending on
traffic volume has reduced congestion by between 16% and 40%, depending on the section [25].

3.2 Road Safety

The effects mentioned in the previous section can also have a positive impact on road safety, as
decreases in the speed limits can lead to a reduction in the speed differences between successive vehicles,
resulting in a decline in rear-end collisions.
Reference [26] presents a simulation-based study showing the potential safety benefits of DSL
using a real-time crash prediction model integrated with a microscopic traffic simulation model. The
study found that dynamic speed limits can reduce average total crash potential by approximately 25%, by
temporarily reducing speed limits during hazardous traffic conditions. Positive effects of DSL have also
been found in other simulation-based studies, such as [27].
Regarding the study of the DSL implemented, an analysis of crash data in Germany has shown
that the use of dynamic speed limit and speed warning signs reduced the crash rate by 20 to 30% [28].
Other German studies cited by [22] estimate a reduction in the number of accidents of over 30% (A5
motorway, near Frankfurt), and a similar decline in fatalities by more than 60%. In Stuttgart, the
reduction in accidents caused by fog conditions is as high as 86%.
In the U.K., [17] analysed data from the M25 in order to compare them with the trends. The
impact of introducing the controlled motorway driving environment (mainly DSL and managed lanes) has
been an estimated reduction in injury accidents of 10% during the period of operation, and a decrease in
the ratio of damage of 20%.
The aforementioned programme in the South of France also had very positive road safety results,
with crashes reduced by 10-20% [25].

Table 1. Summary of evaluation variables used in different research studies

Ref. Case study Variable Effects

Traffic Performance
[5] Simulation of 12 rural roads Standard deviation of the average speed Depending on scenarios
[21] Simulation Traffic volume, travel time, queue length, Variable
number of stops
[13] Simulation Total travel time 21% reduction
[31] A5 Motorway, Germany Intensity 10% increase
[23] Simulation Capacity 2%increase
[17] M25, U.K. Throughput 1.5%increase
[18] E4 Stockholm Lane changes 50%reduction
[25] A7 France Congestion 16-40%reduction
Road Safety
[26] Simulation Total crash potential 25% reduction
[27] Simulation M6, U.K. Time-to-collision Headway Depending on scenarios
[28] German motorways Crash rate 20% reduction
[17] M25, U.K. Damage ratio 20% reduction
[25] A7 France Crash number 10-20% reduction
Emissions
[30] Simulation Total emissions 35% reduction
[17] M25, U.K. Emissions levels Between 2-8%reduction
[17] M25, U.K. Noise levels 0.7-2.3 dB reduction
[9] A12, Austria NO₂ levels 3.6% reduction

3.3 Environmental effects

It is well-known that improved traffic flows can have a significant impact on emission levels [29].
There are very few approaches based on simulating emissions in DSL. Of particular note is the simulation
of a case study based on model predictive control, where the total emissions are reduced by over 35%
[30].
Returning to the case of the M25 motorway in the U.K., in [17] it was found that vehicle emissions
have dropped as a result of reducing start-stop driving. Depending on the particular emissions measured,
the decrease is between 2% and 8%. Fuel consumption has also been reduced. In parallel, there has been a
favourable impact on noise as a result of the introduction of DSL systems between Junction 15 and 16.

120
 The 13th International Conference “RELIABILITY and STATISTICS in TRANSPORTATION and COMMUNICATION - 2013”

The reduction in stop-start driving and the improved compliance with the speed limits have reduced the
weekday traffic noise adjacent to the motorway by around 0.7 decibels, with reductions at some points of
up to 2.3 dB.
Another example can be found in Inn Valley in Austria. The effects of the implementation of DSL
were analysed on this motorway after one year of operation (November 2007 to November 2008). In a
before/after evaluation the results show that NO2 emissions were reduced by 3.6%. Also, the NO2
limit value for short-term exposure (half-hour limit: 200 g/m³) was exceeded only twice during the first
year of operation, while without the DSL in operation, it is estimated that it would have been exceeded
nine times [9].

4. Methodology to Evaluate VSL Systems Using an Aggregate Effectiveness Indicator

4.1 Effectiveness indicator. Definition

Table 1 shows the large number of variables which are used in the scientific literature and other
public reports to evaluate the effects of DSL. This fact highlights the need to find a single variable which
makes possible to evaluate the system’s effectiveness easily and concisely by aggregating the potential
effects on traffic performance, road safety and emissions.
With regard to traffic performance, several of the aforementioned studies point out that the
homogenization of speed (i.e., lower acceleration rates) contributes to a smooth traffic flow [19], an
increase in capacity [20] and the attenuation of shockwaves [13].
Likewise, road safety has been proven to be related with traffic and speed homogeneity [2], [27], [32],
[33]. It can therefore be concluded that there is a clear relation between speed variations and number of
accidents.

Figure 2. Regression fit for CO emissions as a function of speed and acceleration rates. Source [34]

Many research studies [34]-[37] also state that, apart from mean or average speed, positive
acceleration rates also have a major impact on emissions, as shown on Figure 2 from [34].
It has thus been possible to pinpoint instant acceleration as a key factor by evaluating the
effectiveness of implementing DSL, and then proposing an aggregate indicator as follows:
Positive Accumulated Acceleration (PAA) is defined as the sum of the speed variations on a
particular road section.
Mathematically, it is the cumulative integral of the positive acceleration law (1).

If then
(1)

Otherwise

Graphically, PAA is the positive area of the region bounded by the acceleration law, as shown on
Figure 3.

121
Session 3. Transport Systems

Figure 3. Graphic representation of PAA indicator from an acceleration function

The PAA indicator makes it possible to compare the same section before and after the
implementation of DSL, thus evaluating its effectiveness.

4.2 Data collection and evaluation

As already mentioned, the PAA indicator is simply based on speed variations, and the data
required to calculate it is relatively easy to obtain.
Speed data are often collected by induction loops located at certain points on the road, but this
method makes it impossible to establish the speed evolution between two loops, and leads to the
possibility of distorted results.
It is thus essential to obtain speed data at short time intervals, and the methodology proposed is
therefore based on GPS technologies. With a small portable device it is very simple to collect and
download speed data, position and so on every second, which allows a very accurate speed profile to be
obtained along the road section analysed.
The methodology is based on a before & after evaluation, by observing the evolution of the PAA
indicator. Ideally, the number of trip repetitions should be fairly high in order to limit variations caused
by other factors such as meteorology, extraordinary events, incidents, etc. In any case, the trips must be
made in the same time slot and on days with similar behaviour in terms of traffic. In the event of a
limited sample, particular care must be taken to ensure that the conditions are almost the same. The
traffic intensity upstream must be guaranteed to be substantially the same when performing the before
& after trips.
Once the valid data has been processed and selected, the implementation of DSL can be valued
positively if the indicator PAAa (activated) is lower than PAAd (deactivated).

5. Pilot Test Case: West Section of Madrid M30 Motorway

5.1 Description
Madrid is a city of about 3.5 million inhabitants, and up to 6 million in its metropolitan area. The
city is surrounded by three motorway ring-roads, with the M30 the closest to the city centre.
In the afternoon peak hours on a normal working day, the M30 has high traffic levels southbound
on its east and west sections. In an attempt to avoid this habitual congestion and its externalities, the
Madrid Traffic Department is testing a DSL system based on recommended speed limits.
The tested section is a three-lane motorway (southbound) with traffic intensity in the afternoon
peak hours of around 3,300 veh/h. (upstream), and a length of 5.8 km. Most of the section is limited to 90
km/h, except the last 100 m., where the limit is 70 km/h. (tunnel entrance). The congestion is usually
caused by the bottleneck situated at the M500 junction, as around 2,800 vehicles merge into the M30.

122
 The 13th International Conference “RELIABILITY and STATISTICS in TRANSPORTATION and COMMUNICATION - 2013”

The DSL system consists of three Variable Message Signs (VMS) situated before the M500
junction. These VMS display a recommended speed limit of 40, 60 or 80 km/h, depending on the control
algorithm. This is based on instant speed and traffic intensity data recorded by induction loops situated
along the section.

Figure 4. West Madrid motorway ring-road and the 5.8 km. test section (from A to B) with the location of the Variable Message
Signs (VMS). The bottleneck junction where the congestion usually starts (M-500) is highlighted

A microscopic car study was undertaken in the afternoon peak traffic period between 18:00-20:00.
A total of nine trips were made on 6 and 7 June (Tuesday and Wednesday) with the DSL system
activated. One week later (12 and 13 June) another nine trips were performed at exactly the same times,
this time with the DSL system deactivated. The intensity levels upstream for the test days were very
similar, with a maximum deviation of 2.63% from the mean value (Table 2). The weather was sunny and
there were no particular incidents or accidents during the test trips, except for unusual congestion on 6
June, which caused the system not to be automatically activated.

Table 2. Traffic flow intensities upstream (Measuring point PM 22421)

Date
Time Mean Max. mean deviation
05-June 06-June 12-June 13-June
18:00 3159 3196 3114 3324 3198 -2.63%
19:00 3501 3544 3506 3440 3498 -1.65%

The mobile study was carried out using an instrumented vehicle (Skoda Fabia TDI) equipped with
a GPS data recorder (747+ GPS Trip Recorder), which was subsequently downloaded as an Excel Sheet
(.csv) and georeferenced (.kml) documents.
The data collected included travel distance (m), position and speed (m/s), recorded every second,
enabling the PAA to be obtained as defined in the previous section.
Likewise, seven trips (Figure 5) in free flow (southbound mornings) were performed in order to
study the variability of the PAA in similar conditions and to isolate the effects of DSL from any other
which may influence the results (small disturbances and changes in driving style).

123
Session 3. Transport Systems

Figure 5. PAA indicator values in free flow. Obtained in the morning hours of 5 and 6 June

From this analysis, it can be concluded that in free flow and in similar conditions, 99% (confidence
level= 0.99, ά=0.01) of the PAA results present a deviation from the median of less than 2.19 (confidence
limits). Therefore any deviations greater than this value will be assigned to the effects of DSL.

5.1 Analysis of results

Table 3 shows the resulting values of the PAAa and PAAd from the trips performed on 5 and 12
June. The results on Wednesday 6 are considered invalid, as the system was automatically disconnected
due to the unusual and extreme congestion (recorded speed under the operation thresholds).
Table 3. Values obtained for PAA Effectiveness indicator. Congested trips
Activated Deactivated
Tuesday 5 and 12. Afternoon peak
Time PAA values Time PAA values
18:30 91.68 18:29 92.04
18:52 88.20 18:52 88.89
19:11 72.23 19:10 90.99
19:32 59.62 19:32 105.88
19:51 63.54 19:55 69.36
Average 75.05 Average 89.43
Wednesday 6 and 13. Afternoon peak
18:28 125.84 18:27 99.20
18:50 124.06 18:49 86.62
19:14 130.74 19:13 90.68
19:38 73.21 19:38 76.82
Average 113.46 Average 88.33

When the effects of DSL are isolated from any other effects, as described in the paragraphs above,
the result shows that the PAAa falls by an average of 13.1%, compared to the PAAd.

Figure 6. Comparison of the effectiveness indicator (PAA) on Tuesday 5 and 12 June

124
 The 13th International Conference “RELIABILITY and STATISTICS in TRANSPORTATION and COMMUNICATION - 2013”

Figure 6 shows that PAAa and PAAd values are fairly similar, except for the trips that are highly
affected by congestion. An analysis of the speed profiles on Figure 7 and the indicator values shows that
in the 19:11 trip, the speed distribution is more homogeneous, although the congestion levels are similar.
This fact causes the congestion on the following trip (19:32) to remain at similar levels (or even to
decrease) while DSL is activated, and the queue length to increase while deactivated.

Figure 7. Comparison of speed profiles for trips with DSL activated and deactivated

6. Conclusions

After classifying Variable Speed Limits, the literature review has shown that in many cases VSL
(and in particular VSL based on dynamic control) have been beneficial in terms of traffic performance, road
safety and environmental effects. Based on the accumulated acceleration in a section (or instantaneous speed
variations) the methodology described provides a single indicator (PAA) to evaluate whether the
implementation of VSL is working properly and has the potential to produce the desired effects.
To evaluate the feasibility of the methodology on a practical level, a pilot study was carried out on
a stretch of the M30 motorway ring-road in Madrid. This demonstrated the defined PAA effectiveness
indicator to be specific, measurable, reliable and track-able.
Once the effects of driving variability have been statistically bounded by analyzing the trips in free
flow, the variability in traffic intensities requires a greater number of routes. Future research in relation to
this indicator could be directed towards establishing quantitative relationships between changes in the
value of the PAA effectiveness indicator and the VSL effects (increase in capacity, accident rates,
emission levels, etc.).

Acknowledgments
This work was supported in part by the European Commission, under the project ICT-
EMISSIONS, "Development of a methodology and tool to evaluate the impact of ICT measures on road
transport emissions", Grant agreement no: 288568

References

1. Giles, M.J. (2004). Driver speed compliance in Western Australia: a multivariate analysis. Transport
Policy, 11(3), 227-235.
2. N.vanNes, Brandenburg, S. & D. Twisk. (2008). Dynamic speed limits; effects on homogeneity of
driving speed. Intelligent Vehicles Symposium, IEEE, 269.
3. Elvik, R. (2005). Speed and road safety - Synthesis of evidence from evaluation studies,
Transportation Research Record: Statistical Methods; Highway Safety Data, Analysis, and
Evaluation; Occupant Protection; Systematic Reviews and Meta-Analysis, 1908, 59-69.

125
Session 3. Transport Systems

4. Ntziachristos, L. & Z. Samaras. (2000). Speed-dependent representative emission factors for catalyst
passenger cars and influencing parameters. Atmospheric Environment, 34(27), 4611-4619.
5. Makarewicz, R. & M. Gałuszka. (2011). Road traffic noise prediction based on speed-flow diagram.
Applied Acoustics, 72(4), 190-195.
6. Sisiopiku, V. (2011). Variable Speed Control: Technologies and Practice. In Proceedings of the 11th
Annual Meeting of ITS America, Virginia, 2011.
7. Peltola, H. (2000). Seasonally changing speed limits - Effects on speeds and accidents.
Transportation Research Record: Journal of the Transportation Research Board, 1734, 46-51.
8. Wyoming Department of Transport. (May 10, 2012). Variable speed limits replaces seasonal limit
on I-80 section. 2010, Retrieved from
http://www.dot.state.wy.us/wydot/news_info/news_releases;jsessionid=8C31A2CE4E803278FC9A
1703A8425BC4?template=tpl.newsDetail&newsID=910
9. Land Tirol. (May 24, 2012). Tempo 100 auf der Autobahn - warum? (In German). Retrieved from
http://www.tirol.gv.at/themen/verkehr/verkehrsplanung/verkehrsprojekte/tempo100
10. Lehming, B. (2008). Noise reduction plan for Berlin - Action plan. Senats verwaltung für
Gesundheit, Umwelt und Verbraucherschutz. Abt. III Umwelt politik, Referat Immissions schutz,
Berlin.
11. Transport and Main Roads. (2011). School Environment Safety Technical Guidelines. Queensland,
Australia: Queensland Government.
12. Allaby, P., Hellinga, B. & M. Bullock. (2007). Variable Speed Limits: Safety and Operational
Impacts of a Candidate Control Strategy for Freeway Applications. Intelligent Transportation
Systems, IEEE Transactions on, 8(4), 671-680.
13. Hegyi, A., De Schutter, B. & J. Hellendoorn. (2005). Optimal coordination of variable speed limits
to suppress shock waves. IEEE Transactions on Intelligent Transportation Systems, 6(1), 102-112.
14. Xu, J., Liang, F. & W. Yu. (2006). Coordinated Control of Variable Speed Limits Based on Neural
Dynamic Optimization. In Vehicular Electronics and Safety, ICVES, IEEE International Conference
on Vehicular Electronics and Safety, Shanghai, China, 2006 (p. 163).
15. Ghods, A.H., Fu, L. & A. Rahimi-Kian. (2010). An Efficient Optimization Approach to Real-Time
Coordinated and Integrated Freeway Traffic Control. IEEE Transactions on Intelligent
Transportation Systems, 11(4), 873-884.
16. Xiao-Yun, L., Varaiya, P., Horowitz, R., Dongyan, S. & S.E.Shladover. (2010). A new
approach for combined freeway Variable Speed Limits and Coordinated Ramp Metering. In The
13th International IEEE Conference on Intelligent Transportation Systems, Funchal, Portugal,
2010 (p. 491).
17. Highways Agency. (2004). M25 Controlled Motorways. Summary Report. Bristol (UK): Highways
Agency Publications Group.
18. Nissan, A. & X. Bang. (2006). Evaluation of impacts of the motorway control system (MCS) in
Stockholm. In Proceedings of European Transport Conference, AET, Strasbourg, France.
19. García, A. (2009). Estudio sobre la gestión variable de la velocidad en las vías de acceso a las áreas
urbanas. In Cambra Oficial de Comerç, Industria i Navegación de Barcelona. Barcelona, Spain.
20. Heydecker, B.G. & J.D. Addison. (2011).Analysis and modelling of traffic flow under variable
speed limits. Transportation Research Part C: Emerging Technologies, 19(2), 206-217.
21. Zhicai, J., Xiaoxiong, Z. & Y. Hongwei. (2004). Simulation research and implemented effect
analysis of variable speed limits on freeway. In Proc. of the 7th International IEEE Conference on
Intelligent Transportation Systems, Washington, 2004, (p. 894).
22. Schick, P. (2003). Einfluss von Streckenbeeinflussungsanlagen auf die Kapazität von
Autobahnabschnittensowie die Stabilität des Verkehrsflusses: Ph.D. Thesis. Institut für Straßen- und
Verkehrswesen Universität Stuttgart. (In German)
23. Hoogendoorn, S.P. (1999). Multiclass continuum modelling of multilane traffic flow Delft, The
Netherlands: Delft University Press.
24. Knoop, V.L., Duret, A., Buisson, C., & B. van Arem. (2010). Lane distribution of traffic near
merging zones influence of variable speed limits. In Proceedings of the 13th International IEEE
Conference on Intelligent Transportation Systems (ITSC), 2010, (p. 485). Madeira, Portugal.
25. ECMT. (2007). Congestion management measures that release or provide new capacity, Managing
urban traffic congestion, ed. OECD/ECMT, (p. 229). France: OECD Publishing.
26. Lee, C., Hellinga, B. & F. Saccomanno. (2007). Assessing safety benefits of variable speed limits.
Transportation research record: journal of the transportation research board, 1897, 183-190.

126
 The 13th International Conference “RELIABILITY and STATISTICS in TRANSPORTATION and COMMUNICATION - 2013”

27. Piao, J. & McDonald, M. (2008). Safety Impacts of Variable Speed Limits - A Simulation Study. In
Proc. of the 11th International IEEE Conference on Intelligent Transportation Systems, (ITSC),
2008, (p. 833). Beijing, China.
28. Robinson, M. (2000). Examples of Variable Speed Limit Applications. In Speed Management
Workshop, the 79th Annual Meeting of Transportation Research Board, Washington, 2000.
29. Benedekand, C.M., Rilett, L.R. (1998). Equitable traffic assignment with environmental cost
functions. Journal of Transportation Engineering – ASCE, 124, 16–22.
30. Zegeye, S.K., De Schutter, B., Hellendoorn, J. & E.A. Breunesse. (2010). Variable speed limits for
area-wide reduction of emissions. In Proc. of the 13th International IEEE Intelligent Transportation
Systems (ITSC), 2010 (pp. 507-510). Funchal, Portugal.
31. Hoffmann-Leichter. (1997). Untersuchung der Auswirkung der Verkehrsbeeinflussungsanlage auf
der A5 Freidberg – Frankfurt auf Verkehrsablauf, Verkehrssicherheit und Reisezeit, Hessischen
Landes amtes für Straßen- und Verkehrswesen. Falkensee Germany.
32. Fildes, B.N. & S.J. Lee. (1993). The speed review: Road environment, Behavior, Speed limits,
Enforcement and Crashes (pp. 33-34). Monash University, Accident Research Centre.
33. Wegman, F., Aarts, L. & C. Bax. (2008). Advancing sustainable safety: National road safety outlook
for The Netherlands for 2005–2020. Safety Science, 46(2), 323-343.
34. Rakha, V., Van Aerde, M., Ahn, K. & A.A. Trani. (2000). Requirements for evaluating traffic signal
control impacts on energy and emissions based on instantaneous speed and acceleration
measurements. Transportation Research Record: Journal of the Transportation Research Board,
1738(1), 56-67.
35. El-Shawarby, I., Ahn, K. & H. Rakha. (2005). Comparative field evaluation of vehicle cruise speed
and acceleration level impacts on hot stabilized emissions. Transportation Research Part D:
Transport and Environment, 10(1), 13-30.
36. Joumard, R., Jost, P., Hickman, J. & D. Hassel. (1995). Hot passenger car emissions modelling as
a function of instantaneous speed and acceleration. Science of the Total Environment, 169(1-3),
167-174.
37. Ding, Y. & H. Rakha. (2002). Trip-based explanatory variables for estimating vehicle fuel
consumption and emission rates. Water, Air, & Soil Pollution: Focus, 2(5), 61-77.

127