Terraza M, Zhang J, Li Z.

Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

MANEL TERRAZA1 Traffic Engineering

E-mail: manel.terraza@gmail.com Original Scientific Paper
JI ZHANG2 Submitted: 4 Sep. 2020
Email: jzhan121@hawk.iit.edu Accepted: 12 Feb. 2021
ZONGZHI LI, Ph.D.2 DOI: 10.7307/ptt.v33i4.3694
(Corresponding author)
E-mail: lizz@iit.edu
1 Department of Civil Engineering

Universitat Politècnica de Catalunya

Barcelona, Spain
2 Department of Civil, Architectural and Environmental

Engineering
Illinois Institute of Technology
Chicago, IL 60616, USA

INTERSECTION SIGNAL TIMING OPTIMISATION

FOR AN URBAN STREET NETWORK TO MINIMISE
TRAFFIC DELAYS
ABSTRACT 1. INTRODUCTION
The ever-increasing travel demand outpacing avail-
1.1 Background
able transportation capacity especially in the U.S. urban
areas has led to more severe traffic congestion and de- The population growth and economic devel-
lays. This study proposes a methodology for intersection opment have resulted in travel demand increase,
signal timing optimisation for an urban street network especially in urban areas. With the pace of trans-
aimed to minimise intersection-related delays by dy- portation capacity expansion falling significantly
namically adjusting green splits of signal timing plans behind the level of travel demand escalation, many
designed for intersections in an urban street network in urban street networks in the United States are cur-
each hour of the day in response to varying traffic enter- rently operated close to, or even over capacities
ing the intersections. Two options are considered in op- in the AM and PM peak periods. Owning to land
timisation formulation, which are concerned with mini- scarcity in urban areas and prohibitively high costs
mising vehicle delays per cycle, and minimising weighted
of facility delivery, expanding the capacity of an
vehicle and pedestrian delays per cycle calculated using
urban street network is typically not feasible. The
the 2010 Highway Capacity Manual (HCM) method. The
ever-increasing gap of capacity shortage in the
hourly vehicular traffic is derived by progressively exe-
urban street network has then exacerbated recur-
cuting a regional travel demand forecasting model that
could handle interactions between signal timing plans
rent urban traffic congestion in peak periods that
and predicted vehicular traffic entering intersections, is increasingly becoming one of today’s biggest
coupled with pedestrian crossing counts. A computation- challenges. To slow down the deteriorating trend
al study is conducted for methodology application to the of urban traffic conditions, improving the utilisa-
central business district (CBD) street network in Chica- tion efficiency of the existing capacity of an urban
go, USA. Relative weights for calculating weighted ve- street network has been gradually thought of as a
hicle and pedestrian delays, and intersection degrees of possible solution [1, 2]. One of the related mea-
saturation are revealed to be significant factors affecting sures is to reduce delays experienced by individual
the effectiveness of network-wide signal timing optimis- travellers at intersections that are affected by inter-
ation. For the current study, delay reductions are maxi- section signal timing plans deployed in the urban
mised using a weighting split of 78/22 between vehicle area.
and pedestrian delays. In the last several decades, extensive research
has been conducted to develop methods and mod-
KEY WORDS els to estimate intersection delays and identify
urban network; intersection; vehicle delay; pedestrian effective means to mitigate such impacts. For ex-
delay; signal timing; optimisation. ample, Koutsopoulos and Habbal [3] studied how

Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592 579

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

different methods for calculating vehicle delays having some intersection approaches undersaturat-
at intersections affect results obtained from a traf- ed and the remaining approaches oversaturated to
fic assignment model. These researchers used the match real-world situations more closely.
1985 Highway Capacity Manual (HCM) method Mazloumi et al. [9] introduced a formula to cal-
for impacts assessment based on four scenarios culate intersection delays in a traffic assignment
with different levels of details on intersection ge- model addressing impacts of driver behaviour. The
ometry designs and traffic control. It was conclud- formula demands modest data inputs, making it
ed that the 1985 HCM procedure worked well and, very useful for real-life traffic assignments. As the
although increased details of geometric designs HCM hypothesis on intersection delays could not
and traffic control could reduce errors significant- be assumed in Iran, these researchers proposed a
new formula and compared the results and its ad-
ly, the computational intensity became drastically
justments to reality with the HCM model. This for-
increased. Thus, adopting a simple but practical
mula and the HCM model were then introduced to
model for intersection delay estimation would be
a traffic assignment model to generate traffic flow
desirable when conducting a large-scale network
predictions. When compared to field traffic data in
analysis. Fambro and Rouphail [4] proposed an
the city of Mashhad, Iran, a better goodness of fit
improved approach over the 1994 HCM method was obtained from the proposed formula for de-
for calculating intersection delays that formed the lay estimation. It was concluded that the specific
basis of the 2010 HCM method. It incorporates circumstances of different countries would affect
one more term in the delay equation to handle the nature of intersection delays, implying that the
traffic-actuated control and effects of upstream formula for calculating delays be adjusted accord-
signals, along with additional parameters inside ingly.
the equations to enhance the accuracy of delay Wolshon and Taylor [10] analysed changes in
estimation in oversaturated traffic conditions. The intersection delays when applying a time respon-
enhanced model was verified using both field and sive signal control strategy introduced by Lowrie
simulated data. Aashtiani and Iravani [5] devel- [11] as part of assessing the hypothesised effective-
oped a method to calculate vehicle delays at sig- ness of real-time signal control compared with the
nalised intersections in the city of Tehran, Iran. pre-timed control. This analysis was applied in six
Specifically, the Webster equation was integrated intersections in the city of South Lyon, France. To
into the EMME/2 software to iteratively compute facilitate cross comparisons of the two strategies,
the cycle length and green splits using the pre- a formula was introduced to calculate intersection
dicted traffic volumes with model validation by delays. The results showed that vehicle delays are
field traffic counts as inputs, followed by vehicle distributed more evenly across intersection ap-
delay calculation [6, 7]. The main feature of the proaches in the application of the SCATS system.
proposed method is the low amount of input data However, it could lead to increase of total inter-
section delays in some cases. Quiroga et al. [12]
needed, which makes it a decent choice for adop-
also proposed a model for calculating intersection
tion in cities where fine-grained data is not avail-
delays. For the model application, global position-
able. Liu et al. [8] proposed a method for delay
ing system (GPS) devices were installed in several
calculation by categorising intersection traffic into
vehicles to collect precise data on vehicle deceler-
undersaturated, near saturated, and oversaturat-
ation, acceleration, and stopped delay times. These
ed conditions. Each traffic condition was further researchers identified a linear relationship between
separated as the presence and absence of an initial delays and rates of vehicle deceleration and accel-
queue at the intersection. The method first focused eration manoeuvres.
on deriving the delay formula for each of the six Li et al. [13] proposed a new model for comput-
cases and then on an application to three-phase ing intersection delays through image processing.
signalised intersections. Two types of intersection It is the first research that utilised the image pro-
utilisation cases were studied where all intersec- cessing technique to calculate intersection delays
tion approaches were undersaturated and all ap- without having problems of dealing with overlap-
proaches were conversely oversaturated. Further ping images of vehicles and the need for model
refinements were made to cover mixed cases of calibration before being adopted for predictive

580 Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

analysis resembles that of a simulation model. An formulation. This treatment is advantageous in

algorithm was developed to effectively detect and presence of high pedestrian traffic in urban areas
track vehicles using two camera-based detectors, where shared mobility of vehicle and pedestrian
one at upstream of an intersection and the other at traffic should be given duly cognisance in the in-
the stop line. The delay of each vehicle was calcu- tersection signal timing design.
lated as the difference between the time a vehicle
passes the stop line at the downstream intersection 1.2 Motivation
and the time for the vehicle traversing through the
same distance at a free-flow speed. Results were Travellers have experienced benefits of delay
compared with field data to verify the accuracy reductions from intersection signal timing optimi-
and precision of the proposed method, showing sation following various methods gradually refined
the usefulness of the algorithm in real-time appli- over the years. The refinements are manifested by
cation due to low time consumption. the analytical capabilities in spatial and time dimen-
Zhu et al. [14] proposed a real-time network sions concerning the extent of facility coverage and
model for vehicle navigation applications incorpo- time fragmentation of signal timing designs. The fa-
rating intersection delays. The significance of this cility coverage has been enlarged from an isolated
model is in that it not only includes intersection de- intersection, intersections along one or more paral-
lays that many researchers do not consider or treat lel corridors, to intersections of an entire network.
them as a constant if considered, but also re-opti- As for the time fragmentation of signal timing de-
mises vehicle paths as new information becomes signs, it has been expanded from peak and non-peak
available. The model includes an augmented Dijk- periods of a day, hour-by-hour, to virtually sub-
stra algorithm where the optimal results are updat- hour intervals. Delay calculations have also been
ed in response to the additional information. This broadened from considering vehicular delays only
model was tested in Chongqing, China with results to holistically addressing both vehicle and pedes-
of optimal travel paths obtained highly consistent trian delays. For instance, the model developed by
with real-world choices. The model application is Roshandeh et al. [16] is a typical case superior to
potentially limited to transportation agencies capa- many existing methods and models because it con-
ble of collecting, processing, and analysing mas- ducts signal timing optimisation for intersections
sive data for related model applications. in an urban network, considers mobility of both
Qiao et al. [15] proposed a model for estimating vehicle and pedestrian traffic, and derives optimal
intersection delays based on the fuzzy logic theory.
signal timing plans by minimising weighted vehi-
The use of fuzzy logic allows to consider not only
cle and pedestrian delays where the relative weights
technical factors but also non-technical factors
of vehicle delays could be adjusted in the range of
such as weather conditions that affect delays. Fac-
10%–100% with a decrement of 10% in each option
tors incorporated into the proposed model include
of weighting combinations. However, no attempt
flow ratio, green time, cycle time, and weather
was made to identify the best weighting combina-
condition. The model was executed alongside with
the Webster model and the HCM model to esti- tion for vehicle and pedestrian delays in signal tim-
mate intersection delays using field data collected ing design. In this respect, it motivates a more thor-
in Hong Kong, China. The model estimation errors ough investigation into the use of widely adopted
were found to be significantly lower than those of methods for estimating vehicle and pedestrian de-
the Webster and HCM applications. lays such as the HCM method and identification of
Some recent studies have concentrated on op- the specific weighting combination for vehicle and
timisation of intersection signal timing plans for pedestrian delays as the objective to be minimised.
an urban network. For instance, Roshandeh et Intuitively, vehicle delays should have a higher rel-
al. [16] developed a model using the shockwave ative weight as opposed to pedestrian delays since
theory to design signal timing plans for intersec- the mode share of travel by the auto mode and other
tions in an urban street network aimed to minimise private vehicle modes such as carpool, vanpool, and
vehicle delays experienced by travellers. In the other types of ridesharing is typically predominant.
proposed model, both vehicle and pedestrian de- Determining such a weight will help transporta-
lays have been incorporated into the optimisation tion planners and traffic operators design the truly

Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592 581

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

optimal intersection signal timing plans to mitigate saturation flow rate during this period. Saturation
delays encountered by vehicle drivers and pedestri- flow rate is the equivalent hourly rate at which pre-
ans utilising an urban street network. viously queued vehicles can traverse an intersection
approach under prevailing conditions if the green
1.3 Outline signal is always available and no lost times are ex-
perienced. This means that a flow rate higher than
The remainder of this paper is organised as fol-
the saturation flow rate will lead to a queue at the
lows: The next section describes the proposed meth-
end of the cycle. The uniform delay of an intersec-
odology for determining the most favourable share
tion approach can be calculated as:
of weights between vehicle and pedestrian delays
0.5 $ C j $ b 1 - C l
for intersection signal timing optimisation. The sub- gj 2
j T
sequent section focuses on applying the proposed d 1, j, k, T = gj $ Cj (2)
methodology in a computational study. Finally, the 1 - min ^ 1, X j, k h $ C
j

last section presents a study summary and draws

conclusion. where:
Cj – cycle length designed for intersection j to
complete the signal sequence of green,
2. PROPOSED METHODOLOGY
yellow, and red intervals, in sec;
2.1 Method for estimating intersection gj – effective green time available to perform
delays vehicle movements in approach k of
Since its inception, HCM has been widely used intersection j, in sec;
by transportation engineers for traffic operations Xj,k – vj,k/cj,k as the volume-to-capacity ratio
analysis. For this reason, the current study employs indicating how close to saturation in approach
the 2010 HCM method [1] to estimate vehicle and k of intersection j where vj,k and cj,k are traffic
pedestrian delays at individual intersections accord- volume of the critical travel lane of approach
ing to entering traffic and the adopted signal timing k of intersection j, in veh/h/ln, and the
plans. maximum sustainable flow rate at which
vehicles can be expected to drive through the
Estimation of intersection vehicle delays intersection approach, in veh/h/ln;
According to the 2010 HCM method, intersec- min(1, Xj,k) – the lower value between 1 and Xj,k.
tion-related vehicle delays are defined as average
Incremental delay. Incremental delay is the ex-
control delays experienced by all vehicles that ar-
tra delay due to non-uniform arrivals and tempo-
rive at an intersection for the duration of an analy-
rary cycle failures (random delay), as well as de-
sis period. They generally consist of uniform delay,
lay caused by sustained periods of oversaturation.
incremental delay owing to traffic fluctuation, and
initial queue delay caused by vehicles unable to be Typically, the calculation assumes that there is no
released from the intersection in one or more previ- unmet demand causing initial queues at the start
ous signal cycles. of the analysis period. Incremental delay contains
two components: The first portion is associated to
d VEH, j, k, T = d 1, j, k, T + d 2, j, k, T + d 3, j, k, T (1) the fluctuations in demand during the cycle where
where: demand might exceed capacity at some point. The
dVEH,j,k,T – intersection control-related vehicle second portion accounts for delay caused by a sus-
delays associated with approach k of tained oversaturation experienced by an intersection
intersection j during the analysis period approach during the analysis period.
T, in sec/veh, which are comprised of:
d 2, j, k, T 900T <^ X j, k - 1 h + ^ X j, k - 1 h2 + F (3)
8K j $ I j $ X j, k
=
d1, j,k,T – uniform delay; c j, k $ T
d2, j,k,T – incremental delay; where:
d3, j,k,T – initial queue delay. Kj – the incremental delay factor to account for
Uniform delay. Uniform delay is an estimate the effect of signal control type with a value
of delay assuming random uniform arrivals, sta- of 0.5 recommended for pre-timed signals;
ble flow, and no initial queue. It also assumes Ij – the upstream filtering/metering adjustment
one effective green time during the cycle and one factor to handle the effect of metering arrivals

582 Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

for the upstream intersection to intersection j using approach k of intersection j during

with a value of 1.0 suggested for an isolated the cycle length period of Cj, in
intersection. sec/ped/cycle;
Initial queue delay. The equation used for esti- gWALK,j,k – the effective green time for pedestrians
mating incremental delay assumes that there is no crossing approach k of intersection j,
initial queue in presence at the start of the analysis in sec.
period. The initial queue delay accounts for the ad-
ditional delay caused by an existing initial queue. 2.2 Method for network-wide intersection
This queue is the result of unmet demand in the pre- signal timing optimisation
vious period. The initial queue delay of an intersec-
Network-wide intersection signal timing optimi-
tion approach is computed as:
sation could be carried out using two models, which
d 3, j, k, T = v $ T $ e t $
3600
2 2 2
Q b, j, k + Q e, j, k - Q eo, j, k Q e, j, k - Q eo, j, k Q b, j, k o
+ - 2c (4) include a base model focusing on minimising vehi-
2 2c j, k
cle delays per vehicle per cycle, and an enhanced
j, k j, k

with model aimed to minimise the weighted vehicle and

pedestrian delays per traveller per cycle. The ob-
Q b, j, k ^ h h = Q b, j, k ^ h - 1 h + T ^ v j, k ^ h h - c j, k h (5) jective functions of the base and enhanced models
have the following specifications:
Q e, j, k = Q b, j, k + t ^ v j, k - c j, k h (6)
Base model:
Z] T ^ v - c h, t = T, if v j, k > c j, k RS Kj VW
/ b d VEH,j,k,T $ CT j l WWWW
]] j, k j, k SS
Q eo, j, k = []
]] 0, t =
Q b , j, k (7) SS J A (9)
c j, k - v j, k , otherwise Minimise d VEH /cycle = SS / k=1
Kj
WW /J A
\ Sj=1 W
T X
where: Enhanced model:
h – the period of calculation that typically lasts Minimise
d WEIGHTED /cycle = w $ d VEH /cycle + ^ 1 - w h $ d PED /cycle (10)
from one hour to 24 hours of a daily period;
t – the adjusted duration of unmet demand in where:
the analysis period requiring queue dVEH/cycle – the value of vehicle delays
dissipation, in hour; in sec/vehicle/cycle;
Qb,j,k – initial queue of approach k of intersection j dPED/cycle – the value of pedestrian delays
at the start of the analysis period, in sec/pedestrian/cycle;
Qe,j,k – queue at the end of the analysis period; dWEIGHTED/cycle – the weighted value of vehicle
Qeo,j,k– queue at the end of the analysis period due and pedestrian delays,
to oversaturation and without initial in sec/traveller/cycle,
queue, in veh. w – the relative weight assigned to vehicle
In contrast with calculating uniform and incre- delays;
mental delays by lane group that are averaged over j – 1, 2, …, JA, number of intersections in
different lane groups, initial queue delay is estimat- the area A street network;
ed directly for the entire intersection approach. This k – 1, 2, …, Kj, number of approaches of
technical treatment is due largely to lack of infor- intersection j.
mation on the types of initial queues across different The weight w in Equation 10 reflects relative im-
lane groups. portance assigned to vehicle delays to determine op-
timal signal timing plans with the objective to yield
Estimation of intersection pedestrian delays the lowest level of weighted delays per traveller per
According to the 2010 HCM method, the delays cycle. Practically, the choice of a weighting factor
of pedestrians waiting to cross an intersection ap- depends on the traffic management authority’s poli-
proach are computed as follows: cy where relative weights of vehicle and pedestrian
^ C j - g WALK, j, k h2 delays could change by the target area for optimal
d PED, j, k, C j = 2$Cj (8)
signal timing design and by the period of the day.
where: Without loss of generality, the relative importance
dPED,j,k,C – the average pedestrian delay per of vehicle delays in calculating weighted vehicle
j
pedestrian per cycle for pedestrians and pedestrian delays can be altered from 100% to

Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592 583

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

10%. For network-wide signal timing optimisa- intersection arrival time for some vehicular travel-
tion, the base and enhanced models are separately lers. In this respect, the optimal signal timing plans
applied with varying weighting combinations. for the first hour could be used to rerun the region-
Figure 1 illustrates the model execution process. al travel demand model to update vehicular traf-
As seen in the above framework, iterative compu- fic predicted using the initial signal timing plans
tations are conducted progressively from the 1st from the 2nd hour onward until the 24th hour of the
hour to the 24th hour of the daily analysis period day. Next, the updated predictive vehicular traf-
to handle interactions of vehicular traffic only or fic along with pedestrian traffic estimates for the
vehicle and pedestrian traffic jointly expected to second hour could be utilised to refine the initially
enter individual intersections with signal timing designed signal timing plans for the second hour
plans designed in each hour. The vehicular traf- to establish the optimal signal timing plans for the
fic details in terms of autos, trucks, and buses for second hour. Likewise, iterative calculations could
through, left-turn, and right-turn movements along be progressed from the 3rd hour to the 24th hour of
different intersection approach in a second-by-sec- the day to eventually derive network-wide optimal
ond time interval are generated by executing a re- signal timing plans in each hour of the daily anal-
gional travel demand forecasting model that could ysis period.
readily handle interactions of assigned traffic and
travel time with given signal timing plans and
3. COMPUTATIONAL STUDY
their updates. The fine-grained vehicular traffic
details are aggregated into hourly vehicular traffic The street network of the densely populated
volumes, coupled with hourly pedestrian counts Chicago central business district (CBD, Chicago,
extracted from external resources, to derive the USA) is selected to apply the base and enhanced
optimal signal timing plans for each hour of the models for network-wide intersection signal tim-
24-hour period. ing optimisation, respectively. For the enhanced
Using signal timing plans for AM, PM, and rest model application, varying weighting combina-
of day periods currently deployed in the field as in- tions are considered for computing the weighted
puts, the regional travel demand forecasting model vehicle and pedestrian delays. As seen in Figure 2,
could be executed to predict vehicular traffic vol- the study area covers four parts, which are the CBD
umes hour-by-hour for the 24-hour period. Based core of Chicago Loop as subarea 1; north of Loop
on the predictive vehicular traffic and pedestrian as subarea 2; immediate west of Loop as subarea
crossing traffic using all intersections in the net- 3; and near west of Loop as subarea 4 that respec-
work in the first hour, the cycle lengths of signal tively contain 143, 333, 76, and 269 signalised in-
timing plans could be computed. Fixed offsets for tersections, adding to a total of 821 intersections.
adjacent intersections could be determined accord-
ing to their spacing and average vehicle speed in 3.1 Data preparation
between them. To ensure signal coordination, cy-
cle lengths for all intersections in the network are The underlying efforts begin with obtaining
constrained to be identical or with one-half cycle the Chicago regional travel demand forecasting
length. Next, green splits corresponding to various model to help generate predictive vehicular traffic
phases of signal cycles designated to individual utilising intersections in the study area network,
intersections in the network could be adjusted to collecting data on pedestrian crossing counts to de-
achieve the lowest level of vehicle delays per ve- rive pedestrian traffic estimates, and soliciting de-
hicle per cycle or weighted vehicle and pedestrian tails of intersection signal timing plans currently in
delays per traveller per cycle for the entire net- use. Next, MATLAB scripts as outlined in Figure 1
work. The cycle lengths, offsets, and green splits are developed for applying the proposed base and
are essential components of the optimal signal tim- enhanced models for intersectional signal timing
ing plans. optimisation. Finally, the regional demand model
Compared with the initial signal timing plans is interfaced with the computer scripts to facilitate
for the first hour, the optimal signal timing plans executing the two sets of models in succession to
for the first hour achieving the lowest level of de- progressively generate details of predictive ve-
lays per traveller may trigger re-routing or alter the hicular traffic and develop optimal signal timing

584 Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

INITIALISATION

Hour= 0

Node= 0

Link= 0

Detailed traffic predictions in node

Cycle length, green splits, offset

Xj,k = vj,k/cj,k; T=1 hour; Kj= 0.5; Ij= 1

No. of through No. of left turn No. of right turn

lanes > 0? lanes > 0? lanes > 0?

Yes No Yes No Yes No

Compute d1,j,k,T,T Compute d1,j,k,T,LT Compute d1,j,k,T,RT
d1,j,k,T,T = 0 d1,j,k,T,LT = 0 d1,j,k,T,RT = 0
and d2,j,k,T,T for and d2,j,k,T,LT for and d2,j,k,T,RT for
d2,j,k,T,T = 0 d2,j,k,T,LT = 0 d2,j,k,T,RT = 0
through lanes left lanes right lanes

Average d1,j,k,T and d2,j,k,T in approach

Compute queue in hourly period

Compute ta Yes No Compute ta

Compute Qeo v>c? Qeo= 0

Compute Qe

Compute d3
Pedestrian walking counts in node
Link= Link+1
Compute pedestrian delays dPED

Vehicle delays d1,j,k,T, d2,j,k,T, and d3,j,k,T in node j

Pedestrian delays dPED in node

Node= Node+1

Hour= Hour+1

Update d1,j,k,T, d2,j,k,T, d3,j,k,T, and dPED/cycle

Minimise dVEH/cycle, or dWEIGHTED/cycle = w . dVEH/cycle + (1-w) . dPED/cycle

Execute regional travel demand forecasting model

STOP

Figure 1 – Computational process of progressive daily network-wide signal timing optimisation

Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592 585

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

tation network details in terms of connectivity, geo-

metrics, and operations control of highway network
nodes and links; transit routes, stops and stations, and
operational schedules in the metropolitan area; and
Immediate pedestrian walkways in the CBD area.
West of Loop North of Loop
(Subarea 3) (Subarea 2) Calibration and validation of the Chicago mod-
el were successfully completed in 2012 using field
data on traffic volumes, speeds, and travel times by
travel lane collected from over 800 continuous free-
way traffic counting stations in the metropolitan area
Near and 1,200 urban intersection counting locations in
West of Loop Loop
(Subarea 4) (Subarea 1)
the city of Chicago, USA. Approximately one-half
of field data was used for model calibration involv-
ing adjustments of global and local parameters of
route choice and driver behaviour models and refine-
ments of origin-destination (O-D) trips in the Chica-
go metropolitan area for model fine-tuning to match
Figure 2 – Illustration of the Chicago CBD street network for predicted traffic with field traffic data collected on
intersection signal timing optimisation link-specific volumes, speeds, and travel times. The
remaining portion of field data was utilised for mod-
plans for intersections in the study area network to
el validation via quantitative assessment of model
achieve minimised delays per traveller per cycle
results using such measures as Geoffrey E. Havers
by hour of the day.
(GEH) values and confidence intervals, as well as a
Predictive vehicular traffic data qualitative evaluation of reasonableness checks on
Data on hourly vehicular traffic predictions is top O-D travel paths assigned to travellers. For more
essential to implement the base and enhanced mod- details, see Li et al. [17].
els in the proposed methodology for network-wide
signal timing optimisation. For each round of signal Pedestrian count data
timing optimisation, the Chicago model developed Pedestrian walking counts are crucial in applying
based on the TRANSIMS toolbox as the next gen- the enhanced model to develop optimal signal tim-
eration tool for transportation planning, traffic oper- ing plans with the objective of minimising weighted
ations, and emergency management analyses in the vehicle and pedestrian delays. Pedestrian count data
Chicago metropolitan area is employed to progres- associated with the study area comes from a pedes-
sively derive high fidelity data on vehicular traffic trian traffic study administered by the city of Chica-
predictions aggregated into data on hourly predictive go in 2007. Relying on 510 counting locations, the
vehicular traffic utilising intersections in the study study counted over three million pedestrians walking
area network. in either direction of sidewalks within the study area
As background information, the Chicago model from 7:45 AM to 5:45 PM for 10 hours. An annual
covers over 10,000 square miles of Chicago met- pedestrian traffic growth rate of one percent is used to
ropolitan area in northeastern Illinois, USA that update pedestrian counts to keep consistency with the
contains 1,962 traffic analysis zones (TAZs). In a time period of the forecasted vehicular traffic.
typical weekday, 28.5 million person-trips are pro-
duced, which could be classified by travel purposes Initial signal timing plan data
of Home-Based Work (HBW), Home-Based Other For the 821 intersections in the study area, details
(HBO), Non-Home-Based (NHB), airport trips, ex- of intersection signal timing plans initially used in the
ternal trips, light truck trips, mid-sized truck trips, field are obtained from the city of Chicago, USA.
and large-sized truck trips with corresponding hourly
diurnal distributions. In the CBD area, 3.64 million 3.2 Model application
daily person-trips are made, equivalent to 12.8%
of total trips in the metropolitan area. The Chicago MATLAB scripts for executing the signal
model also contains multimodal regional transpor- timing optimisation process

586 Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

MATLAB scripts are developed to apply the iter- pending on the subareas being selected for inter-
ative computational process of signal timing optimis- section signal timing optimisation. If only select-
ation as outlined in Figure 1. The scripts are concerned ing one subarea for network-wide signal timing
with modules for data handling, delay calculation, optimisation, subarea 2 or 4 is the best candidate
and signal timing optimisation. The data handling which could achieve delay reductions by 4% or
module imports and exports data on vehicle vol- 3%. With multiple subareas considered, delay re-
umes from and to the Chicago model, and inter- ductions are achieved if both subareas 2 and 4 are
section signal timing plans before and after opti- included into signal timing optimisation, the op-
misation, as well as importing data on pedestrian tions are subareas 2 and 4, subareas 1, 2, and 4,
traffic for analysis. The delay calculation module subareas 2, 3, and 4, and subareas 1, 2, 3, and 4.
conducts vehicle delay estimation using the base When network-wide signal timing optimisation
model, and weighted vehicle and pedestrian delay is conducted using the enhanced model, the effec-
estimation using the enhanced model as Equations 9 tiveness of delay reductions is highly sensitive to
and 10, respectively. The signal timing optimisation combinations of relative weights assigned to vehi-
module renders adjustments of green splits of sig- cle and pedestrian delays. Intersection signal tim-
nal timing plans to minimise the level of average ing optimisation is effective in delay reductions
delays for travellers utilising intersections in the when relative weights of vehicle delays fall in the
study area. range of 40%–80%, most preferably in the range
Results of 70%–80%. With one, two, three, and four subar-
Table 1 lists average hourly vehicle volumes eas chosen for network-wide signal timing optimi-
entering signalised intersections in the street net- sation, the highest levels of delay reductions could
works of each subarea and the entire study area reach 14%, 13%, 11%, and 9%, respectively.
before and after network-wide signal timing opti-
Discussion
misation, along with hourly vehicle volumes as-
The analysis results are not unexpected. Sub-
sociated with several signalised intersections in
area 1 is commonly called the Chicago Loop that
the subarea 2 network as labelled in Figure 1. The
stands as the largest employment centre in the en-
average daily vehicle volumes per intersection
tire Chicago metropolitan area, which attracts and
after network-wide signal timing optimisation
produces extensive trips to and from the subarea,
reduced by 3.5–7.6% for signalised intersections
leading to high demand of external–internal, in-
in the street networks of subareas 1, 2, and 4, and
increased by 6.0% for signalised intersections in ternal–external, and internal–internal vehicle and
subarea 2 network. Whereas the average daily ve- pedestrian travel. Subarea 3 is located west of sub-
hicle volumes per intersection after network-wide area 1 bounded by I-94 in the west and Chicago
signal timing optimisation for all signalised inter- River in the east. The east-west arterial streets of
sections in the entire study area network remain subarea 3 not only accommodate external–internal,
to be steady, which decreased marginally by 2.0%. internal–external, and internal–internal vehicular
The findings suggest that network-wide signal tim- traffic for the subarea but also serve as the feed-
ing optimisation would lead to traffic redistribu- er roads for vehicular traffic entering and exiting
tion from one subarea network to the other, while subarea 1. This creates high volumes of vehicular
keeping the total trips using the entire study area traffic within the subarea. Compared with intersec-
stable. tions in subareas 2 and 4, the degrees of saturation
Table 2 summarises computational study results. of intersections in subareas 1 and 3 are potentially
The effectiveness of signal timing optimisation is higher because of higher volumes of vehicle and
measured by percentage reductions in vehicle de- pedestrian traffic especially in subarea 1. The ve-
lays measured in seconds per vehicle per cycle, or hicle-to-vehicle, vehicle-to-pedestrian, and pedes-
reductions in weighted vehicle and pedestrian de- trian-to-pedestrian interactions in the two subareas
lays measured in seconds per traveller per cycle. are also more intensive. This leaves less room to
Negative values are indicative of positive effects further mitigate delays at intersections in the two
on delay reductions because of network-wide in- subareas utilising signal timing optimisation. Con-
tersection signal timing optimisation. For the base versely, the comparatively lower volumes of vehi-
model application, mixed results are obtained de- cle and pedestrian traffic in subareas 2 and 4 cou-

Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592 587

 Table 1 – Average hourly vehicle volumes entering intersections in the subarea and entire study area before and after network-wide signal timing optimisation [veh/hr]

588
Subarea 1 Subarea 2 (333 Ints) Subarea 3 Subarea 4 Study area
(143 Ints) Int 1 Int 2 … Int 333 Subarea (76 Ints) (269 Ints) (821 Ints)

Bef Aft Bef Aft Bef Aft … Bef Aft Bef Aft Bef Aft Bef Aft Bef Aft
1 377 383 299 295 49 50 … 705 706 207 206 236 235 108 107 239 239
2 440 447 346 346 51 54 … 834 833 244 244 269 267 122 120 277 277
3 445 448 360 362 71 65 … 818 835 241 242 279 278 127 126 282 282
4 459 466 362 362 52 52 … 924 927 248 248 286 285 128 127 289 290
5 562 572 364 352 53 51 … 910 915 301 301 358 356 156 154 356 357
6 944 954 737 738 119 111 … 1589 1604 493 493 584 583 241 240 585 586
7 1641 1620 862 872 105 127 … 2081 2189 789 820 1117 1073 447 424 1039 1025
8 2447 2374 1219 1248 153 167 … 2606 2798 1137 1188 1788 1711 754 707 1594 1562
9 2911 2817 1279 1310 174 182 … 2997 3353 1372 1457 2281 2196 1038 967 1978 1948
10 3189 3061 1621 1687 194 180 … 3104 3436 1524 1598 2541 2476 1212 1129 2198 2165
11 3142 3069 1886 2090 252 279 … 2976 3291 1507 1625 2617 2516 1296 1167 2223 2196
12 2937 2836 1903 1968 246 236 … 2884 3211 1476 1615 2490 2255 1266 1110 2122 2041
13 2903 2810 1953 2191 268 212 … 2963 3218 1481 1602 2410 2250 1274 1130 2086 2031
14 2969 2833 1960 2244 233 241 … 2926 3266 1442 1609 2486 2336 1264 1150 2115 2074
15 3129 2960 1837 2120 257 253 … 2941 3445 1500 1648 2613 2441 1318 1177 2218 2155
16 3313 3203 1991 2285 206 227 … 2894 3038 1610 1671 2729 2658 1328 1250 2334 2299
17 3502 3278 1862 2290 269 262 … 3062 3481 1608 1741 2836 2700 1357 1255 2412 2353
18 3342 3105 1656 2075 182 151 … 2962 3316 1501 1627 2848 2677 1345 1235 2354 2274
19 2576 2479 1506 1757 139 163 … 2818 3274 1224 1306 2132 2011 1031 940 1810 1763
20 1750 1743 1187 1110 188 193 … 2586 2364 904 877 1326 1329 646 642 1199 1189
21 1054 1020 918 973 106 103 … 1480 1480 511 520 713 701 330 328 671 663
22 600 591 643 667 52 61 … 970 986 316 319 398 392 178 173 386 382
23 459 453 344 343 59 58 … 926 935 246 249 282 280 135 132 288 286
24 439 433 333 334 42 44 … 895 894 239 241 270 265 125 123 276 273
Daily 45533 43956 27428 30019 3520 3522 49851 53795 22122 23448 35890 34268 17225 15912 31331 30711
Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

Diff. -3.5% 9.5% 0.1% 7.9% 6.0% -4.5% -7.6% -2.0%

Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

Table 2 - Effectiveness of network-wide signal timing optimisation for different subarea combination scenarios

Reductions in average delays

Subarea Enhanced model (Relative weight of vehicle delays, w)
Base model
90% 80% 70% 60% 50% 40% 30% 20% 10%
1 7% 15% -4% 0% 5% 7% 8% 10% 11% 12%
2 -4% 7% -14% -10% -8% -7% -5% -2% -1% 21%
3 4% 10% -10% -6% -2% 0% 2% 6% 7% 6%
4 -3% 7% -7% -5% -5% -3% -1% 4% 6% 3%
12 0% 9% -11% -7% -4% -3% -1% 1% 3% 19%
13 6% 13% -6% -2% 2% 5% 6% 9% 10% 10%
14 0% 10% -6% -4% -1% 0% 2% 6% 8% 6%
23 -2% 8% -13% -10% -7% -6% -4% -1% 1% 18%
24 -3% 7% -11% -8% -7% -5% -3% 1% 2% 13%
34 -1% 8% -8% -6% -4% -2% 0% 4% 6% 3%
123 0% 10% -11% -7% -4% -2% -1% 2% 3% 17%
124 -1% 8% -9% -7% -4% -3% -1% 2% 4% 13%
134 1% 10% -7% -4% -2% 0% 2% 6% 8% 6%
234 -2% 7% -11% -8% -6% -5% -3% 1% 3% 12%
1234 -1% 9% -9% -6% -4% -3% -1% 3% 4% 12%

pled with street networks with similar capacities ing pedestrian delay considerations will likely cause
offer a greater potential for reducing intersection excessive pedestrian delays, especially for subarea
delays in the two subareas. 2 with high volumes of pedestrian traffic.
Comparing the two models for network-wide Figures 3a and 3b present intuitive results of delay
signal timing optimisation, the base model could reductions after signal timing optimisation for all
be considered a special case of the enhanced model possible subarea combinations and most desirable
by assigning the relative weight of vehicle delays subarea combinations using the base model and the
as 100% in the calculation of weighted vehicle and enhanced model, respectively.
pedestrian delays. Although subareas 2 and 4 are As seen in Figure 3a, signal timing optimisation is
desirable candidates to deploy signal timing optimi- not guaranteed to be effective in achieving intersec-
sation, the use of the base model is less favourable tion delay reductions for all possible combinations
than the enhanced model in that the optimal signal of subareas in the study area if minimising vehicle
timing plans developed by the base model exclud- delays is the objective as the base model to develop
Reductions in average delays at intersections [%]

Reductions in average delays at intersections [%]

25 1 25 2
2
20 3 20
4 4
15 15
12
10 13 10
14 24
5 23 5
24 124
0 34 0
123
-5 -5
124 234
-10 134 -10
234
-15 1234 -15 1234
100 90 80 70 60 50 40 30 20 10 100 90 80 70 60 50 40 30 20 10

Relative weight assigned to vehicle delays [%] Relative weight assigned to vehicle delays [%]

a) All possible subarea combinations b) Most desirable subarea combinations

Figure 3 – Graphic representation of the effectiveness of network-wide intersection signal timing optimisation

Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592 589

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

the optimal signal timing plans. By simultaneously 38% to 20%, signal timing optimisation is expect-
considering vehicle and pedestrian delays and striv- ed to trigger steady delay increases by up to 7%.
ing for achieving the minimum level of weighted When further decreasing relative weights of vehicle
delays as the objective of the enhanced model, the delays from 20% to 10%, signal timing optimisation
effectiveness of signal timing optimisation is sen- becomes ineffective again and the delay increases
sitive to weighting combinations for vehicle and could exceed 20%.
pedestrian delays. For all possible subarea combi-
nations, signal timing optimisation is generally ef- 4. SUMMARY AND CONCLUSION
fective when the relative weights assigned to vehi-
cle delays are in the range of 70%–85%. The extent This study has proposed a methodology for net-
of reductions in delays measured in seconds per work-wide intersection signal timing optimisation
traveller per cycle could be as high as 15%. When over a 24-hour daily period. The objective is to
relative weights of vehicle delays are assigned to achieve the lowest level of delays for each travel-
be higher than 85%, signal timing optimisation is ler by dynamically altering green splits of signal
likely to increase delays by 5%–15%. When rela- timing plans designed for individual intersections
tive weights of vehicle delays decrease from 70% in the network hour-by-hour in response to vary-
to 20%, signal timing optimisation is expected to ing vehicle and pedestrian traffic entering the in-
trigger steady increases of delays by up to 11%. tersections. Two models have been introduced for
When relative weights of vehicle delays are further delay calculation in the optimisation formulation,
decreased from 20% to 10%, signal timing optimi- which aim to minimise vehicle delays per driver
sation becomes ineffective for all possible subarea and minimise weighted vehicle and pedestrian de-
combinations with which the highest level of delay lays per traveller calculated according to the 2010
increases could exceed 20%. HCM method. The relative weights for vehicle
Figure 3b shows the effectiveness of network-wide and pedestrian delays can be adjusted in the range
signal timing optimisation for the most desirable of 10%–100%. The hourly vehicular traffic is de-
subarea combinations, which include options of rived by progressively executing a regional trav-
subarea 2, subarea 4, subareas 2 and 4, subareas 1, el demand forecasting model capable of handling
2, and 4, subareas 2, 3, and 4, and subareas of 1, interactions between dynamically predicted traffic
2, 3, and 4 of the study area combined. In compar- and intersection signal timing plans. Predictions of
isons with findings depicted in Figure 3a, a similar vehicular traffic entering approaches of individual
but more consistent trend is revealed. In general,
intersections in the network are carried out in al-
signal timing optimisation is effective in achieving
ternation with the design of signal timing plans in
intersection delay reductions for all desirable sub-
consecutive time intervals. This helps derive opti-
area combinations if vehicle delays are minimised
mal signal timing plans by the hour of the day and
as the base model. The percentage reductions in
obtain high-fidelity predictive vehicular traffic on
vehicle delays measured in seconds per vehicle
a second-by-second basis that could be aggregated
per cycle could reach 2–4%. With weighted vehi-
cle and pedestrian delays minimised according to into hourly predictive vehicular traffic. MATLAB
the enhanced model, signal timing optimisation scripts have been created to facilitate methodology
is effective when the relative weights assigned to application.
vehicle delays are in the range of 38–85%. More- A computational experiment targeting the Chi-
over, the highest levels of delay reductions rang- cago CBD street network has been conducted for
ing from 7.5% to 15% are achieved for all desir- network-wide intersection signal timing optimisa-
able subarea combinations using a weighting split tion. The study area is generally classified into four
of 78% and 22% for vehicle and pedestrian delays. subareas that contain 821 signalised intersections.
When relative weights of vehicle delays vary from Of the two models proposed for delay calculations,
85%–98%, signal timing optimisation is likely to the base model that computes vehicular delays
increase delays where the worst case occurs when only is essentially a special case of the enhanced
the relative weight of vehicle delays is assigned as model that calculates weighted vehicle and pedes-
92%, leading to an increase in delays by 8%. When trian delays by assigning a relative weight of 100%
relative weights of vehicle delays decrease from to vehicular delays.

590 Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

According to model execution results, delay DATA AVAILABILITY

reductions are found to be sensitive to relative Some data, models, or code that support the find-
weights assigned to vehicle and pedestrian delays ings of this study are available from the correspond-
calculated in the objective function of the en- ing author upon reasonable request.
hanced model for network-wide intersection sig-
nal timing optimisation. Also, not all four subareas ACKNOWLEDGMENTS
are effective in achieving delay reductions. This
The authors are grateful for the assistance of
seems to suggest that network-wide signal timing
transportation agencies in the Chicago metro-
optimisation could be potentially effective if de-
politan area for data collection, processing, and
grees of saturation of intersections in a street net-
computational analysis. Additional thanks are ex-
work are comparatively low where the remaining
tended to publication support of NNSFC 71871029,
intersection capacities could still accommodate a
CHD 300102219306, and CHD 300102218404.
significant amount of traffic. In this case, fine-tun-
ing of signal timing plans for intersections in one MANEL TERRAZA1
or more subarea networks could help improve the Correo electrónico: manel.terraza@gmail.com
efficiency of intersection capacity utilisation with JI ZHANG2
delay reductions. Correo electrónico: jzhan121@hawk.iit.edu
The application of the enhanced model with ZONGZHI LI, Doctor en Filosofía2
(Autor correspondiente)
vehicle and pedestrian delays calculated using the Correo electrónico: lizz@iit.edu
2010 HCM method for network-wide intersection 1 Departamento de Ingeniería Civil

signal timing optimisation has confirmed that rel- Universitat Politècnica de Catalunya
ative weights assigned to the two categories of de- Barcelona, España
2 Departamento de Ingeniería Civil, Arquitectónica
lays as the objective to be minimised, and degrees y Ambiental Instituto de Tecnología de Illinois
of saturation of intersections in the study area net- Chicago, IL 60616, EE.UU.
work are two factors affecting the effectiveness of
delay reductions. Network-wide signal timing op- OPTIMIZACIÓN DEL TIEMPO DE LA SEÑAL
timisation is not always effective in reducing de- DE INTERSECCIÓN PARA UNA RED
lays for all possible subarea combinations. Thus, DE CALLE URBANA PARA MINIMIZAR
the proposed methodology could be used by mu- LAS RETRASOS DEL TRÁFICO
nicipalities to identify alternative subarea combi-
RESUMEN
nations potentially effective in delay reductions
La creciente demanda de viajes que supera la ca-
from network-wide signal timing optimisation and
pacidad de transporte disponible, especialmente en las
to prioritise them for real-world deployments ac-
áreas urbanas de los EE. UU., Ha provocado retrasos y
cordingly. congestiones de tráfico más graves. Este estudio propone
The computational study has revealed that una metodología para la optimización del tiempo de las
the proposed methodology is useful in identify- señales de intersección para una red de calles urbanas
ing subareas that might benefit from intersection con el objetivo de minimizar los retrasos relacionados
signal timing optimisation for delay deductions, con las intersecciones ajustando dinámicamente las di-
quantifying the benefits, and prioritising the best visiones verdes de los planes de tiempo de las señales
option for field deployments. However, extensive diseñados para las intersecciones en una red de calles
efforts are needed in data collection, preparation, urbanas en cada hora del día en respuesta a la variación
and computation, especially iterative analysis of del tráfico. entrar en las intersecciones. Se consideran
vehicular traffic predictions, delay estimation, and dos opciones en la formulación de optimización, que se
refieren a minimizar las demoras de vehículos por ciclo
signal timing optimisation. Therefore, the pro-
y minimizar las demoras ponderadas de vehículos y pea-
posed methodology may be adopted for real-world tones por ciclo calculadas usando el método del Manual
applications by large municipalities with rich data de Capacidad de Carreteras (HCM) de 2010. El tráfi-
resources of travel demand, and traffic operations co vehicular por hora se deriva mediante la ejecución
and control, as well as budget strength for field progresiva de un modelo de pronóstico de demanda de
deployments that typically require upgrading and viajes regional que podría manejar las interacciones en-
replacing hardware and software of intersection tre los planes de tiempo de señales y el tráfico vehicular
control. previsto que ingresa a las intersecciones, junto con los

Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592 591

 Terraza M, Zhang J, Li Z. Intersection Signal Timing Optimisation for an Urban Street Network to Minimise Traffic Delays

recuentos de cruces de peatones. Se realiza un estudio [8] Liu GP, Zhai RP, Pei YL. A calculating method of in-
computacional para la aplicación de la metodología a tersection delay under signal control. Proceedings of the
la red de calles del distrito comercial central (CBD) en 2007 IEEE Conference on Intelligent Transportation
Systems, 30 Sep. - 3 Oct. 2007, Bellevue, WA. New York:
Chicago, EE. UU. Se revela que los pesos relativos para
Institute of Electrical and Electronics Engineers (IEEE);
calcular las demoras ponderadas de vehículos y pea- 2007. p. 648-653. DOI: 10.1109/ITSC.2007.4357783
tones, y los grados de saturación en las intersecciones,
[9] Mazloumi E, Moridpour S, Mohsenian H. Delay function
son factores importantes que afectan la efectividad de la for signalized intersections in traffic assignment models.
optimización del tiempo de señal en toda la red. Para el Urban Planning and Development. 2010;136(1): 67-74.
estudio actual, las reducciones de retrasos se maximizan DOI: 10.1061/(ASCE)0733-9488(2010)136:1(67)
utilizando una división de ponderación de 78/22 entre [10] Wolshon B, Taylor WC. Analysis of intersection delay
retrasos de vehículos y peatones. under real-time adaptive signal control. Transportation
Research Part C: Emerging Technologies. 1999;7(1): 53-
72. DOI: 10.1016/S0968-090X(99)00011-X
PALABRAS CLAVE
[11] Lowrie PR. The Sydney coordinated adaptive traffic
red urbana; intersección; retraso del vehículo; retraso system- principles, methodology, algorithms. Proceed-
de peatones; temporización de la señal; mejoramiento. ings of the International Conference on Road Traffic
Signaling, 30 Mar. - 1 Apr. 1982, London, United King-
dom. London: Institution of Electrical Engineers; 1982.
REFERENCES
p. 67-80. Available from: https://pascal-francis.inist.fr/
[1] TRB. Highway capacity manual. Transportation Research vibad/index.php?action=getRecordDetail&idt=PAS-
Board. Washington, D.C.: National Academic Press; 2010. CAL82X0330853 [Accessed 6th Mar. 2020].
[2] Li Z. Transportation asset management: Methodology and [12] Quiroga CA, Bullock D. Measuring control delay at
applications. First Edition. Boca Raton, FL: CRC Press, signalized intersections. Transportation Engineering.
Taylor and Francis Group; 2018. 1999;125(4): 271-280. DOI: 10.1061/(ASCE)0733-947X
[3] Koutsopoulos HN, Habbal M. Effect of intersection de- (1999)125:4(271)
lay modeling on the performance of traffic equilibrium [13] Li Z, Li N, Liu F. An effective calculating method of
models. Transportation Research Part A: Policy and signalized intersection delay. Proceedings of the 2009
Practice. 1994;28(2): 133-149. DOI: 10.1016/0965- International Conference on Digital Image Processing,
8564(94)90034-5 7-9 March 2009, Bangkok, Thailand. New York: Institute
[4] Fambro DB, Rouphail NM. Generalized delay model of Electrical and Electronics Engineers (IEEE); 2009. p.
for signalized intersections and arterial streets. Trans- 211-215. DOI: 10.1109/ICDIP.2009.41
portation Research Record. 2007;1572: 112-121. DOI: [14] Zhu Z, et al. A simplified real-time road network mod-
10.3141/1572-14 el considering intersection delay and its application on
[5] Aashtiani HZ, H. Iravani H. Use of intersection delay vehicle navigation. Applied Mechanics and Materials.
functions to improve reliability of traffic assignment mod- 2011;58-60: 1959-1965. DOI: 10.4028/www.scientific.
els. Proceedings of the 14th Annual International EMME/2 net/AMM.58-60.1959
Conference, 20-22 October 1999, Chicago, USA. Mon- [15] Qiao F, Yi P, Yang H, Devarakonda S. Fuzzy logic based
treal: INRO Consultants; 1999. p. 20-28. Available from: intersection delay estimation. Mathematical and Com-
https://www.academia.edu/2605461/Use_of_intersec- puter Modelling. 2002;36(11/13): 1425-1434. DOI:
tion_delay_functions_to_improve_reliability_of_traffic_ 10.1016/S0895-7177(02)00298-4
assignment_model [Accessed 11th Feb. 2020]. [16] Roshandeh AM, et al. New methodology for intersection
[6] Webster FV. Traffic signals settings. Road Research Lab- signal timing optimisation to simultaneously minimise
oratory Crowthorne, London, United Kingdom. Report vehicle and pedestrian delays. Transportation Engineer-
number: 39, 1958. Available from: https://www.sinalde- ing. 2014;140(5): 04014009. DOI: 10.1061/(ASCE)
transito.com.br/artigos/traffic_signals_webster.pdf [Ac- TE.1943-5436.0000658
cessed 2nd Mar. 2020]. [17] Li Z, et al. Development and application of the TRAN-
[7] INRO. EMME/2 user’s manual. Software Release 7. SIMS toolbox for transportation operations management
INRO Consultants; 1994. Available from: https://www. in and around Chicago central area. Federal Highway
inrosoftware.com/en/products/emme/ [Accessed 5th Mar. Administration, U.S. Department of Transportation.
2020]. Phase I Final Report, 2012.

592 Promet – Traffic&Transportation, Vol. 33, 2021, No. 4, 579-592