Durham Research Online
Deposited in DRO:
27 May 2008
Version of attached le:
Published Version
Peer-review status of attached le:
Peer-reviewed
Citation for published item:
Papageorgiou, M. and Diakaki, C. and Dinopoulou, V. and Kotsialos, A. and Wang, Y. (2003) 'Review of road
trac control strategies.', Proceedings of the IEEE., 91 (12). pp. 2043-2067.
Further information on publisher's website:
http://dx.doi.org/10.1109/JPROC.2003.819610
Publisher's copyright statement:
c
2003 IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for
advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists,
or to reuse any copyrighted component of this work in other works must be obtained from the IEEE.
Additional information:
Use policy
The full-text may be used and/or reproduced, and given to third parties in any format or medium, without prior permission or charge, for
personal research or study, educational, or not-for-prot purposes provided that:
•
a full bibliographic reference is made to the original source
•
a link is made to the metadata record in DRO
•
the full-text is not changed in any way
The full-text must not be sold in any format or medium without the formal permission of the copyright holders.
Please consult the full DRO policy for further details.
Durham University Library, Stockton Road, Durham DH1 3LY, United Kingdom
Tel : +44 (0)191 334 3042 | Fax : +44 (0)191 334 2971
https://dro.dur.ac.uk
Review of Road Traffic Control Strategies
MARKOS PAPAGEORGIOU, FELLOW, IEEE, CHRISTINA DIAKAKI, VAYA DINOPOULOU,
APOSTOLOS KOTSIALOS,
AND YIBING WANG
Contributed Paper
Traffic congestion in urban road and freeway networks leads to
a strong degradation of the network infrastructure and accordingly
reduced throughput, which can be countered via suitable control
measures and strategies. After illustrating the main reasons for
infrastructure deterioration due to traffic congestion, a comprehen-
sive overview of proposed and implemented control strategies is
provided for three areas: urban road networks, freeway networks,
and route guidance. Selected application results, obtained from
either simulation studies or field implementations, are briefly
outlined to illustrate the impact of various control actions and
strategies. The paper concludes with a brief discussion of future
needs in this important technical area.
Keywords—Driver information, freeway network control, intelli-
gent transportation systems, ramp metering, route guidance, traffic
signal control, urban network control.
I. INTRODUCTION
A. Traffic Congestion and the Need for Traffic Control
Transportation has always been a crucial aspect of
human civilization, but it is only in the second half of
the last century that the phenomenon of traffic congestion
has become predominant due to the rapid increase in the
number of vehicles and in the transportation demand in
virtually all transportation modes. Traffic congestion ap-
pears when too many vehicles attempt to use a common
transportation infrastructure with limited capacity. In the
best case, traffic congestion leads to queueing phenomena
(and corresponding delays) while the infrastructure capacity
(“the server”) is fully utilized. In the worst (and far more
typical) case, traffic congestion leads to a degraded use
of the available infrastructure (reduced throughput), thus
contributing to an accelerated congestion increase, which
leads to further infrastructure degradation, and so forth.
Traffic congestion results in excess delays, reduced safety,
and increased environmental pollution. The following
Manuscript received December 6, 2002; revised July 18, 2003.
The authors are with the Technical University of Crete, Dynamic
Systems and Simulation Laboratory, GR-73100 Chania, Greece (e-mail:
markos@dssl.tuc.gr).
Digital Object Identifier 10.1109/JPROC.2003.819610
impressive statement is included in the European Commis-
sion’s “White Paper—European Transport Policy for 2010”:
“Because of congestion, there is a serious risk that Europe
will lose economic competitiveness. The most recent study
on the subject showed that the external costs of road traffic
congestion alone amount to 0.5% of Community GDP.
Traffic forecasts for the next 10 years show that if nothing is
done, road congestion will increase significantly by 2010.
The costs attributable to congestion will also increase by
142% to reach
80 billion a year, which is approximately
1% of Community GDP.”
The emergence of traffic (i.e., many interacting vehicles
using a common infrastructure) and subsequently traffic
congestion (whereby demand temporarily exceeds the
infrastructure capacity) have opened new innovation needs
in the transportation area. The energy crisis in the 1970s,
the increased importance of environmental concerns, and
the limited economic and physical resources are among the
most important reasons why a brute force approach (i.e.,
the continuous expansion of the available transportation
infrastructure) cannot continue to be the only answer to the
ever increasing transportation and mobility needs of modern
societies. The efficient, safe, and less polluting transporta-
tion of persons and goods calls for an optimal utilization
of the available infrastructure via suitable application of a
variety of traffic control measures. This trend is enabled by
the rapid developments in the areas of communications and
computing (telematics), but it is quite evident that the effi-
ciency of traffic control directly depends on the efficiency
and relevance of the employed control methodologies. This
paper provides an overview of advanced traffic control
strategies for three particular areas: urban road networks,
freeway networks, and route guidance and information
systems.
B. The Control Loop
Fig. 1 illustrates the basic elements of a control loop. The
traffic flow behavior in the (road or freeway or mixed) traffic
network depends on some external quantities that are classi-
fied into two groups.
0018-9219/03$17.00 © 2003 IEEE
PROCEEDINGS OF THE IEEE, VOL. 91, NO. 12, DECEMBER 2003 2043
Fig. 1. The control loop.
• Control inputs that are directly related to corre-
sponding control devices (actuators), such as traffic
lights, variable message signs, etc.; the control in-
puts may be selected from an admissible control re-
gion subject to technical, physical, and operational
constraints.
• Disturbances, whose values cannot be manip-
ulated, but may possibly be measurable (e.g., de-
mand) or detectable (e.g., incident) or predictable
over a future time horizon.
The network’s output or performance is measured via suit-
able indices, such as the total time spent by all vehicles in the
network over a time horizon. The task of the surveillance is
to enhance and to extend the information provided by suit-
able sensors (e.g., inductive loop detectors) as required by
the subsequent control strategyand the human operators. The
kernel of the control loop is the control strategy, whose task
is to specify in real time the control inputs, based on avail-
able measurements/estimations/predictions, so as to achieve
the prespecified goals (e.g., minimization of total time spent)
despite the influence of various disturbances. If this task is
undertaken by a human operator, we have a manual control
system. In an automatic control system, this task is under-
taken by an algorithm (the control strategy). The relevance
and efficiency of the control strategy largely determines the
efficiency of the overall control system. Therefore, when-
ever possible, controlstrategiesshouldbedesignedwithcare,
via application of powerful and systematic methods of opti-
mization and automatic control, rather than via questionable
heuristics [1]. Traffic control strategies for urban road and
freeway networks is the main focus of this overview paper.
C. A Basic Property
For the needs of this paper we will use a discrete-time
representation of traffic variables with discrete time index
and time interval (or sampling time) .A
traffic volume or flow
(in veh/h) is definedas the number
of vehicles crossing a corresponding location during the time
period
, divided by T. Traffic density (in
veh/km) is the number of vehicles included in a road segment
of length
at time kT, divided by . Mean speed (in
Fig. 2. A traffic network.
km/h) is the average speed at time of all vehicles included
in a road segment.
We consider a traffic network (Fig. 2) that receives de-
mands
(in veh/h) at its origins and we
define the total demand
We as-
sume that
, is independent of any control
measures taken in the network. We define exit flows
at
the network destinations
and the total exit flow
We wish to apply control mea-
sures so as to minimize the total time spent
in the network
over a time horizon K, i.e.,
(1)
where
is the total number of vehicles in the network at
time k. Due to conservation of vehicles we have
(2)
hence
(3)
Substituting (3) in (1) we obtain
(4)
The first two terms in the outer sum of (4) are independent
of the control measures taken in the network; hence, mini-
mization of
is equivalent to maximization of the following
quantity:
(5)
Thus, minimization of the total time spent in a traffic net-
work is equivalent to maximization of the time-weighted exit
flows. In other words, the earlier the vehicles are able to exit
the network (by appropriate use of the available control mea-
sures) the less time they will have spent in the network.
D. Traffic Congestion Revisited
The abovebasic property may be used to explain and quan-
tify the degradation of the network infrastructure caused by
traffic congestion, as well as to demonstrate via simple math-
ematics the enormous potential of improvement via suitable
traffic control measures.
As an example, we consider (Fig. 3) two cases for a
freeway on-ramp, (a) without and (b) with metering control.
2044 PROCEEDINGS OF THE IEEE, VOL. 91, NO. 12, DECEMBER 2003
Fig. 3. Two cases, (a) without and (b) with ramp metering; grey
areas indicate congestion zones.
Fig. 4. Two cases, (a) without and (b) with ramp metering.
Let be the upstream freeway flow, be the ramp demand,
be the mainstream outflow in presence of congestion,
and
be the freeway capacity. It is well known that
the outflow
at the head of the congestion is lower by
some 5%–10% than the freeway capacity
. Note that
(else the congestion would not have been
created). In Fig. 3(b) we assume that ramp metering may be
used to maintain capacity flow on the mainstream, e.g., by
using the control strategy ALINEA (see Section III-C). Of
course, the application of ramp metering creates a queue at
the on-ramp but, because
is greater than (increased
outflow!), ramp metering leads to a reduction of the total
time spent (including the ramp waiting time). It is easy to
show that the amelioration
(in %) of the total time
spent is given by
(6)
For example, if
(i.e., the total demand ex-
ceeds the freeway capacity by 20%) and
(i.e., the capacity drop due to the congestion is 5%) then
% results from (6), which illustrates the level of
achievable improvement.
In addition, we consider (Fig. 4) two cases of a freeway
stretch that includes an on-ramp and an off-ramp, namely,
(a) without and (b) with metering control. In order to clearly
separate the different sources of degradation, we will assume
Fig. 5. Detrimental effects of overspilling queues in urban road
networks.
here that , i.e., no capacity drop due to conges-
tion. Defining the exit rate
as the portion of
the upstream flow that exits at the off-ramp, it is easy to show
that the exit flow without control is given by
(7)
while with metering control we have
(8)
Because
holds (else the conges-
tion would not have been created), it follows that
is less
than
; hence, ramp metering increases the outflow, thus
decreasing the total time spent in the system. It is easy to
show that the amelioration of the total time spent in this case
amounts to
(9)
For example, if the exit rate is
, then the ameliora-
tion is
%. If several upstream off-ramps are covered
by the congestion in absence of ramp metering (which is typ-
ically the case in many freeways during rush hours) then the
amelioration achievable via introduction of ramp metering is
accordingly higher.
Summing up these effects in a freeway network, an overall
amelioration of total time spent by as much as 50% (i.e.,
halving of the average journey time) may readily result (see
Section III-D) due to the increased throughput enabled by
ramp metering application.
Similar effects may be observed in saturated signal-
controlled urban traffic networks (Fig. 5). A saturated link
prevents the traffic movements at the upstream intersection
to cross, even though they have the right of way (green
signal). This is a waste of resources (waste of green time)
that contributes to an accelerated increase of congestion due
to vehicles trapped in the upstream links, which leads to
blocking of further upstream intersections, increased waste
of green time, and so forth [2]. This vicious circle frequently
leads to gridlocks in network cycles with devastating effects
for the traffic flow in extended urban areas.
PAPAGEORGIOU et al.: REVIEW OF ROAD TRAFFIC CONTROL STRATEGIES 2045
Fig. 6. Example of signal cycle.
Fig. 7. Cycle time and lost times.
The outlined phenomena make clear that the extended
congestion encountered daily in modern freeway and urban
road networks are not only due to excessive demand ex-
ceeding the network capacity. As a matter of fact, demand
may temporarily exceed the capacity of specific links
leading to limited congestion. The infrastructure degrada-
tion, however, caused by the initially limited congestion,
leads to an unstable escalation when no suitable control sys-
tems are employed to counter this devastating evolution. In
conclusion, the observed extended (in both space and time)
congestion in modern metropolitan areas is indeed triggered
by a temporarily and locally excessive demand, but it is
expanded and maintained due to the lack of suitable control
actions that would prevent the corresponding infrastructure
degradation.
II. R
OAD TRAFFIC CONTROL
A. Basic Notions
Traffic lights at intersections is the major control mea-
sure in urban road networks. Traffic lights were originally in-
stalled in order to guarantee the safe crossing of antagonistic
streams of vehicles and pedestrians; with steadily increasing
traffic demands, it was soon realized that, once traffic lights
exist, they may lead (under equally safe traffic conditions) to
more or less efficient network operations, hence there must
exist an optimal control strategy leading to minimization of
the total time spent by all vehicles in the network.
Although the corresponding optimal control problem may
be readily formulated for any road network, its real-time so-
lution and realization in a control loop like the one of Fig. 1
faces a number of apparently insurmountable difficulties.
• The red–green switchings of traffic lights call for
the introduction of discrete variables, which renders
the optimization problem combinatorial.
• The size of the problem for a whole network is
very large.
• Many unpredictable and hardly measurable
disturbances (incidents, illegal parking, pedestrian
crossings, intersection blocking, etc.) may perturb
the traffic flow.
• Measurements of traffic conditions are mostly
local (via inductive loop detectors) and highly noisy
due to various effects.
• There are tight real-time constraints, e.g., de-
cision making within 2 s for advanced control
systems.
The combination of these difficulties renders the solution of
a detailed optimal control problem infeasible for more than
one intersection. Therefore, proposed control strategies for
road traffic control introduce a number of simplifications of
different kinds or address only a part of the related traffic
control problems. Unfortunately, most proposed simplifica-
tions render the corresponding control strategies less suitable
to address traffic saturation phenomena.
An intersection consists of a number of approaches and
the crossing area. An approach may have one or more lanes
buthas a unique, independent queue. Approaches are used by
corresponding traffic streams (veh/h). A saturation flow
is
the average flow crossing the stop line of an approach when
the corresponding stream has right of way (r.o.w.), the up-
stream demand (or the waiting queue) is sufficiently large,
and the downstream links are not blocked by queues. Two
compatible streams can safely cross the intersection simulta-
neously, else they are called antagonistic.Asignal cycle is
one repetition of the basic series of signal combinations at
an intersection; its duration is called cycle time
.Astage (or
phase) is a part of the signal cycle, during which one set of
streams has r.o.w. (Fig. 6). Constant lost (or intergreen) times
of a few seconds are necessary between stages to avoid inter-
ference between antagonistic streams of consecutive stages
(Fig. 7).
There are four possibilities for influencing traffic condi-
tions via traffic lights operation.
2046 PROCEEDINGS OF THE IEEE, VOL. 91, NO. 12, DECEMBER 2003
