Transportation Research Part A 118 (2018) 52–67

Contents lists available at ScienceDirect

Transportation Research Part A

journal homepage: www.elsevier.com/locate/tra

A comparative study on the operational performance of four-leg

T
intersections by control type
⁎
Abubeker N. Arshia, Wael K.M. Alhajyaseenb, , Hideki Nakamuraa, Xin Zhanga
a
Graduate School of Environmental Studies, Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8603, Japan
b
Qatar Transportation and Traffic Safety Center, College of Engineering, Qatar University, P.O. Box 2713, Doha, Qatar

A R T IC LE I N F O ABS TRA CT

Keywords: The selection of proper control types at intersections at the planning stage is an important
Two-way stop control concern considering the difficulty in comparing the performance of different control types.
Roundabout Therefore, several manuals proposed selection charts to guide traffic engineers in identifying the
Signalized intersection control strategy that yields to the lowest delay and the highest level of service. However, most of
Intersection vehicle delay
these methodologies consider only total major and minor street traffic flows. They do not ne-
Turning traffic
Pedestrians
cessarily suggest to investigate the impact of turning traffic and pedestrian flows, which limits the
applicability of their recommendations. Thus, this study aims to propose a procedure to assess the
selection of the optimum control strategies among two-way stop, roundabout and signalization
using average intersection vehicle delay as the evaluation criterion. The impact of turning traffic
on the major and minor streets as well as the impact of pedestrians are integrated into the de-
veloped analytical methodology for the estimation of average intersection vehicle delay. A ty-
pical layout of isolated four-leg two-way intersection is assumed for the model development. It is
concluded that turning traffic and pedestrians play essential roles in the performance of inter-
sections. As pedestrian flow increases, signal control becomes prevailing over the other control
types.

1. Introduction

Intersections have a significant role in the operation of road networks. They are key players in determining the amount of traffic
that flows in road links. Their capacities are the dominant factor in defining the capacity of the whole road network. Hence, the
efficiency of these spots contributes significantly towards the efficiency of whole road network, as they are the main bottlenecks in
the system. Traffic control at intersections is a complex process where features related to the efficiency such as capacity and delay are
important considerations along with safety and geometrical constraints. Therefore, the availability of rational procedure at the
planning stage for the selecting of control type is advantageous. Some manuals such as HCM (2000) and Japan Roundabout Manual
(2016) proposed selection charts for the control type at intersections, however, the applicability of these recommendations in the real
world is limited since they do not consider turning traffic movements and pedestrian flows.
Typically, intersections are usually controlled by anyone of the followings: regulatory signs, such as two-way stop control (TWSC)
and all-way stop control (AWSC); signalization (SIG) and roundabout (RAB). In Japan, there has recently been a growing interest in
adopting RABs for the intersection control due to their remarkable efficiency at low to medium traffic flows and their positive safety

⁎
Corresponding author.
E-mail addresses: [email protected] (W.K.M. Alhajyaseen), [email protected] (H. Nakamura),
[email protected] (X. Zhang).

https://doi.org/10.1016/j.tra.2018.08.039
Received 10 October 2017; Received in revised form 26 August 2018; Accepted 31 August 2018
0965-8564/ © 2018 Elsevier Ltd. All rights reserved.
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

performance. However, to identify the most efficient control type that yields to the lowest intersection vehicle delay under specific
vehicle and pedestrian flows at the major and minor streets, it is essential to quantify and compare their performances. Few studies
formulated recommendations for the selection of the appropriate control strategies using average control delay as a common measure
of effectiveness. Most of these studies did not consider in detail the impacts of turning traffic and pedestrian flows, which significantly
limited the applicability of their recommendations. It is important to mention that delay is not the only factor to be considered for the
selection of appropriate control types at intersections. Other factors such as safety also need to be considered. However, it is common
that capacity and average intersection vehicle delay are the most determinant factors.
Therefore, this study investigates the traffic demand combination under which each of TWSC, RAB and SIG best performs using
average intersection vehicle delay as a common measure of effectiveness and assuming undersaturation condition for all control
types. Through the analysis, the study examines the impacts of turning traffic and pedestrian flows on the performance area of each
control type. The estimated delay does not include pedestrian delay. The impact of pedestrian demand on vehicle movement and the
setting of intersection control, which may result in significant additional delay to vehicular traffic, is investigated. The consideration
of pedestrian on vehicle operational performance of control types is the unique contribution of this study. Moreover, although few
existing studies addressed the impact of turning traffic, they did not examine this impact in various scenarios of different right and
left turning volume ratios. Furthermore, this study develops charts for the selection of intersection control types that can assist
engineers in the identification of the optimum control type under combinations of major and minor street flows, turning traffic ratios,
and pedestrian flows.
The remaining parts of this paper are organized as follows. In the literature review, relevant studies on comparing the operational
performance of different control types are discussed. Then the methodology for developing the analytical model and the scenario
setting is presented. This is followed by sensitivity analyses on the impacts of turning traffic and pedestrian flows on the estimated
average intersection vehicle delay. Then charts for the selection of intersection control types are presented and discussed. The paper
ends with conclusions and future works.

2. Literature review

Few studies compared the performance of various intersection control types to draw recommendations on their feasibility areas.
For instance, Manual of Uniform Traffic Control Device MUTCD (2009) provided traffic signal warrants, which served as the guiding
principles for the installation of signals at intersections. It stated that the selection and the use of traffic control signals should be
based on an engineering study of roadways, traffic, and other conditions. Furthermore, the manual suggested to control intersections
with TWSC or AWSC in case a traffic signal is not warranted. However, the manual did not provide any recommendation on the
implementation of RAB. Additionally, the impacts of turning traffic and pedestrian flows are not included.
Kyte et al. (1997) proposed an intersection control selection criteria based on the peak hour intersection volume. They compared
TWSC, AWSC, and SIG by using average delay and average queue length as measures of effectiveness. They stated that the analysis
result reasonably matches peak hour volume warrants of MUTCD (1988). They proposed Fig. 1(a) as a quick method for determining
optimum intersection control type. However, RAB was not considered at that time and they did neither investigate the impacts of
turning traffic nor pedestrian flows.
Highway Capacity Manual HCM (2000) provided a selection chart for intersection control types of future facilities as shown in
Fig. 1(b). This exhibit compares TWSC, AWSC, and SIG control based on the peak hour two-way volumes on the major and minor
streets. It did not consider the effects of turning traffic and pedestrian flows. Additionally, the exhibit did not cover RAB as one of the
potential control types.
Yoshioka et al. (2008) compared the performance of TWSC, RAB and SIG. They suggested Fig. 2 as a selection criterion of control
type at intersections for the future planned facilities. This figure is currently referred to in Japan Roundabout Manual (2016) as an
example of the control type selection criteria. Unlike the above studies, their study considered RAB as one of intersection control

Fig. 1. Recommendations on the optimum intersection control types based on minimum average delay.

53
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

Fig. 2. Optimum intersection control types recommended by Japan RAB Manual (2016).

types. However, they applied Adams delay model for RAB although it underestimates the delay at medium and high traffic flows
(Troutbeck and Brilon, 2008). Furthermore, their analysis is based on the assumption that all the traffics are through only at both
major and minor streets. Moreover, the impact of pedestrians on intersection performance was not considered.
Lee et al. (2008) studied the performance of different control types at isolated intersections based on control delay under various
demand levels. They employed HCM (2000) methodologies for estimating average control delay of TWSC, AWSC and SIG. They
hypothesized a four-leg intersection with two-way two-lane (TWTL) streets for the analysis and concluded that the proposed selection
chart by HCM (2000), Fig. 1(b), is inconsistent with their analysis results. On the base of minimizing delay alone, they stated that if
the traffic flow is low on the minor street, TWSC should be used. However, if the traffic flow on the minor street is low to medium,
AWSC is preferred; otherwise, SIG should be favored. Even though they demonstrated the effect of left turning traffic LT (right-hand
driving system) they ignored the impact of right turning traffic (RT) as well as pedestrians. Additionally, they did not include RAB as
a control type.
Prakash et al. (2014) compared the operational efficiency of TWSC, AWSC, RAB and SIG under a range of traffic conditions in
Australia with different volume ratios and turning ratios using SIDRA software. They assumed constant heavy vehicle ratio of 10% in
all scenarios. They concluded that RT movement (left-hand driving system) affects negatively the operational efficiency of TWSC and
AWSC intersections as well as RAB; meanwhile LT movement has a positive impact on their operational efficiency. Furthermore, they
concluded that TWSC and AWSC outperform other control types at low traffic flow while at moderate traffic flow RAB performs best.
At high traffic flow, SIG performs better than other control types. The main limitation of this study is that authors did not consider the
impact of pedestrian movements. Furthermore, they did not present comparison charts between the control types.
In a recent study, Bumjon et al. (2017) utilized HCM (2010) procedures to compare average control delays of TWSC, AWSC, RAB,
and SIG. They assumed an isolated right angle four-leg intersection for the analysis. They considered the impact of left turning traffic.
However, they ignored the impacts of RT (right-hand driving system) traffic and pedestrian flows. Furthermore, they analyzed two
groups of scenarios, one with fixed cycle length of 60 sec for signalized intersections and another with optimized cycle length. Fig. 3
presents the developed charts by Bumjon et al. (2017) showing the performance areas of control types using signal optimization
assuming no turning traffic (Fig. 3a) and 20% of left turning traffic (Fig. 3b). They found that the optimized SIG and RAB outperform
TWSC in all cases (Fig. 3). This contradicts with previous studies (Parkash et al., 2014; Japan RAB Manual, 2016). This is attributed to
assumed low total lost time (8 sec) which underestimates the actual delay by SIG. Another reason can be the missing geometric delay
component in the RAB delay estimation methodology. On the other hand, Parkash et al. (2014) yielded to insignificant impact of LT
traffic (up to 20% of total traffic) on the performance areas. Moreover, they stated that the delay of RAB increases rapidly as the
traffic flow increases where SIG starts to perform better.
In general, previous studies neglected the impact of pedestrians on the operational performance of intersections. Furthermore,
they did not provide clear assessment on the impacts of different turning traffic ratios on the major and minor streets separately.

3. Methodology

3.1. Hypothetical intersection layout

A typical isolated four-leg right angle intersection is hypothesized. The layouts of TWSC, RAB and SIG are demonstrated in Fig. 4.
The directions of traffic movements are defined assuming left-hand driving system where vehicles drive on the left side of the road.
This is similar to the driving system in Japan, United Kingdom and Australia. As shown in Fig. 4(a), exclusive RT lanes are used at all
approaches of the SIG intersection. This layout is adopted since the performance of signal control is highly sensitive to RT traffic
volume (left-hand driving system). Furthermore, adding exclusive RT lane does not require large space since it is usually done by
utilizing the median space and narrowing through lanes. This is why it is common to have exclusive RT lanes at SIG intersections. In

54
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

Fig. 3. Performance areas of different control types with signal optimization proposed by Bumjon et al. (2017).

order to have a comparable layout, exclusive RT lanes are also utilized at TWSC intersections. Since only single lane RAB is assumed
in this study, there is only one lane on each approach of RAB. Furthermore, the width of each connecting road is assumed as 9 m for
all the intersection control types. For RAB, a 30 m inscribed circle diameter (D) and a 5 m circulatory roadway width are assumed. All
these parameters are illustrated in Fig. 4.

3.2. Intersection control parameters

TWSC is an unsignalized intersection where the minor streets are controlled by STOP signs, whereas the major streets are not
controlled and have absolute right of way. Thus, vehicles at the minor streets have to wait and find acceptable gaps between vehicles
on the major streets. At RABs, vehicles at the entry approaches give the right of way to vehicles on the circulatory roadway. Thus, all
approaches are controlled by either yield or stop signs. Drivers on the entry approach need to wait for acceptable gaps between
circulating vehicles. Therefore, gap parameters are the main control parameters for TWSC and RAB. In this study, the proposed gap
parameter values by HCM 6th edition (2016) for TWSC are adopted for vehicles while those proposed by Japan RAB Manual (2016)
and Kang and Nakamura (2015) for RAB are adopted. Tables 1 and 2 list the assumed gap parameters for TWSC and RAB, respec-
tively. Furthermore, it is assumed that all approaches of the RAB are controlled by yield signs.
For SIGs, a simple pre-timed signal control with two possible types of phasing are assumed as shown in Table 3. In the two-phase
plan (SIG-2), RT movements are permitted, thus they have conflicts with opposing through (TH) and LT traffic. Whereas in the four-
phase plan (SIG-4), RT movements are operated with protected phases. Pedestrian movements are shown as dashed lines in the
phasing plans in Table 3. Furthermore, assumed yellow and all-red intervals are also defined in Table 3. At each traffic flow com-
bination (per movement per approach), optimum cycle length is estimated for SIG-2 and SIG-4 by using Webster formula (Webster,
1966) as shown in Eq. (1). The green times for the phases are estimated proportional to their critical flow ratios. Then average
intersection vehicle delay is estimated for each phasing plan.

1.5∗L + 5
Copt =
1− ∑ yi (1)

55
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

Fig. 4. Hypothetical intersections and lane configurations.

Table 1
Gap parameters for TWSC intersections.
Vehicle movement Critical gap tc (sec) Follow-up time tf (sec) Reference

Major RT 4.1 2.2 HCM 6th edition (2016)

Minor LT 6.2 3.3
Through traffic 6.5 4.0
RT 7.1 3.5
With pedestrians impact for all movements 6.2 2.9 Kang and Nakamura (2015)

Table 2
Gap parameters for RAB.
Traffic Condition Critical gap tc (sec) Follow-up time tf (sec) Minimum headway τ (sec) Reference

Vehicle Without pedestrian impact 4.1 2.9 2.1 Japan RAB Manual (2016)
With pedestrian impact 6.2 2.9 1.0 Kang and Nakamura (2015)

where Copt is optimum cycle length (sec); L is total lost time per cycle (sec) and yi is critical flow ratio of phase i. It is important to
mention that the saturation flow rate for each lane is estimated based on the procedure proposed by HCM 6th edition (2016) in which
the impact of turning traffic, lane width and the presence of pedestrians are considered. The assumed base saturation flow rate and
gap parameters are listed in Table 4. Furthermore, not only the optimum cycle length is considered in this study, but also the
minimum cycle length based on minimum vehicle green time and minimum pedestrian green time. Minimum vehicle green time for
each phase is ensured for the practicality and safety requirements. Japan Manual on Traffic Signal Control Revised Edition MTSCRE
(2006) proposes minimum vehicle green times of 15 sec for TH movements on major streets and 5 sec for protected RT movements on

56
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

Table 3
Considered phasing plans and signal timings for SIG.

Table 4
Base saturation flow rate and gap parameters for SIG.
Parameters Movement type Reference

RT TH LT

Base saturation flow rate (veh/h) 1800 2000 1800 MTSCRE, Japan (2006)
Critical gap tc (sec) 4.5 – – HCM 6th edition (2016)
Follow-up time tf (sec) 2.5 – –
Start-up lost time (sec) 2.0

minor streets. For simplification, this study assumes that the phases of the TH movements at the major and minor streets have an
equal minimum green time of 15 sec while exclusive RT phases on the major and minor streets have an equal minimum green time of
5 sec.
Moreover, pedestrian green time requirement is verified and secured for each phase. Minimum pedestrian green time (tp) is
estimated by using Eq. (2) which was proposed by MTSCRE (2006). This formula is similar to that proposed by HCM 6th edition
(2016).
Lp p
tp = +
Vp W ∗sp (2)

where tp is minimum green time for pedestrians (sec); p is the number of waiting pedestrians at the onset of pedestrian green per cycle
(ped/cycle); Lp is crosswalk length (m); Vp is pedestrian crossing speed (assumed as 1 m/sec as proposed by MTSCRE, 2006); W is
crosswalk width (m); and sp is pedestrian discharge flow rate per meter width of the crosswalk (ped/m/sec). Following MTSCRE
(2006), the value of sp is assumed as 0.92 ped/m/sec, which is the proposed value for intersections located in urban areas with
business activities.

3.3. Control delay models

HCM 6th edition (2016) defines vehicle delay as an extra travel time consumed in transportation facility in comparison to the time
consumed when there are no controls and no impact of other road users. Accordingly, the total delay is the sum of different com-
ponents of delay caused by control, traffic, geometry, incidents, and others. Control delay is the delay caused when a traffic
movement reduces speed or stopes due to a traffic sign or a control device. It is commonly used as a measure of effectiveness (MOE)
since it can be directly related to the driver’s perception of the quality of service.
The existing analytical delay models are compared with the HCM delay models. For example, in the case of unsignalized in-
tersections Adams delay model (1963) is compared with HCM 6th edition (2016) delay model. It is observed that Adams delay model
underestimated the delay under high traffic demand. For signalized intersection, Webster (1966) delay model is compared with the

57
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

HCM 6th edition (2016) delay model. Even though the models have the same uniform delay component, it is found that Webster
model underestimated the intersection vehicle delay (in accordance with Andrzej et al., 2001). This can be attributed to the Webster
random delay component which may yield to low values of intersection vehicle delays at low degree of saturation when compared to
HCM time dependent delay model. Therefore, HCM time dependent delay model is utilized in this study, which is consistent with the
delay models of TWSC and RAB. Other than macroscopic analytical delay models, there are several micro-simulations software that
can be used to estimate vehicle delay for different intersection control types. However, these models require many inputs and need
careful calibration for different intersection control types, which is a complicated and time consuming while there is no guarantee to
have more accurate results than the macroscopic models. Furthermore, since this is a comparative study between different inter-
section control types, we have decided to use macroscopic delay models, which are usually used by practitioners to identify proper
control strategies at intersections.

3.3.1. TWSC delay model

At TWSC intersections, the proportion of delay attributed to the stop control aspect on the intersections is referred to as control
delay. It includes the delay due to deceleration to stop at the back of the queue from free flow speed, the move-up time within the
queue, the stopped delay at the front of the queue, and the delay due to acceleration back to free flow speed at the exit approach.
HCM 6th edition (2016) stated that the average vehicle control delay for any minor movement is a function of the capacity of the
approach and the degree of saturation. For 15-minute analysis period, HCM 6th edition (2016) proposed Eq. (3) to estimate vehicle
control delay. It is important to mention that the last constant term in Eq. (3), which is 5 sec/veh, accounts for the deceleration time
of the vehicle from free flow speed to the speed of the queue and the acceleration time of vehicles from the stop line to the free flow
speed.

d=
3600
⎡
v
+ 900T ⎢ x −1 +
v 2
⎛ x −1⎞ +
( )( ) ⎤⎥ + 5
3600
c
vx
c
c ⎢c ⎝c ⎠ 450T ⎥
⎢
⎣ ⎥
⎦ (3)

where d is vehicle control delay (sec/veh); c is movement capacity (veh/h); T is analysis period (assumed as 15 min) and vx is flow
rate of movement x (veh/h). The capacity (c) of each movement of the TWSC intersection is calculated following the gap acceptance
model proposed by HCM 6th edition (2016).

3.3.2. RAB delay model

Entry, circulating and exiting vehicle flows as well as pedestrians are the major traffic flows at RABs. The circulating flow passing
in front of one entry approach and the exiting flow are composed by the entry flows from other approaches. Accordingly, the
operation of RAB is significantly affected by the entry capacity. The estimation of entry capacity in this study is based on the gap
acceptance model proposed by the Japan RAB Manual (2016). For the estimation of RAB average vehicle control delay, and to be
consistent with the utilized delay models for other control types, the proposed formula by HCM 6th edition (2016) is used (Eq. (4)).
However, the last term (dg) in Eq. (4) is added in this study to account for the geometric delay caused by RAB, which is not considered
in the proposed procedure in HCM 6th edition (2016). It is noted that the first three terms in the RAB average vehicle control delay
formula (Eq. (4)) are similar to those of TWSC intersection (Eq. (3)) except minor differences in the third term to account for the yield
control on the subject entry approach of the RAB.

d=
3600
⎡
v
+ 900T ⎢ +
v 2
⎛ −1⎞ +
( )( ) ⎤⎥ + 5∗min ⎛ v , 1⎞ + dg
3600
c
v
c
c ⎢c ⎝ c ⎠ 450T ⎥ ⎝c ⎠
⎢
⎣ ⎥
⎦ (4)

where d is vehicle control delay (sec/veh); c is entry capacity (veh/h); v is entry flow (veh/h) and dg is geometric delay (sec/veh).
Geometric delay (dg) is the delay caused by the geometric layout of the RAB since all vehicles need to turn onto the circulatory
roadway to the desired exit. dg is estimated as the difference between the travel time through the RAB from the moment when
deceleration begins at the entry approach until the time when the vehicle speed reaches the free-flow speed at the exit approach, and
the travel time between the entry and exit approaches when there is no RAB or any other controls. Both travel times are estimated
under the free-flow conditions. As mentioned in the definition of dg, acceleration and deceleration delays due to RAB geometry are
considered. Accordingly, Eq. (5) are adopted from the study conducted by McDonald, et al. (1984) to estimate dg. The model was
developed based on extensive observations on the public roads in the UK.

V −V l V −V l + lc 2 + la ⎫
dg = ⎧ A C + c1 + E C ⎫−⎧ d
⎨
⎩ αd VC αa ⎬
⎭ ⎨
⎩ VA ⎬
⎭ (5)

where dg is geometric delay (sec/veh); VA is entry approach speed (assumed as 25 km/h for urban RAB with one circulatory lane); VE
is exit approach speed (25 km/h); VC is circulatory speed (m/sec); αd is deceleration rate (m/sec2); αa is acceleration rate (m/sec2); ld
is deceleration distance (m); la is acceleration distance (m); lC1 is circulatory distance with the effect of RAB (m) which it is assumed
as πD/2 and 3πD/4 for TH and RT vehicles respectively; lC2 is circulatory distance without the effect of RAB (m) which is assumed as
D and πD/4 for through and RT vehicles respectively; and D is inscribed circle diameter (m). For the estimations of αa, αd, and VC, the
developed models by McDonald et al. (1984), which are shown in Eqs. (6), (7) and (8) respectively, are used.

58
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

1.11(VE−VC )
αa = + 0.02
VE (6)

1.06(VA−VC )
αd = + 0.023
VA (7)

Vc for TH and LT movements = 0.96 D + 2.03

Vc for RT movement = 0.84( R entry approach + R exit approach ) (8)

3.3.3. SIG delay model

Average control delay experienced by vehicles at SIG intersection consists of three components as stated in the HCM 6th edition
(2016) and expressed in Eq. (9).
d = d1 + d2 + d3 (9)
where d is vehicle control delay (sec/veh); d1 is uniform delay assuming uniform arrival (sec/veh); d2 is incremental delay which
considers the effect of random arrival and cycle failure during the analysis period (sec/veh); and d3 is initial queue delay which
accounts for the delay due to presence of initial queue at the start of analysis period (sec/veh). This study assumes unsaturated
conditions for all intersection control types. Thus, the impact of initial queue delay due to continuous overflow can be neglected.
However, it is important to note that at high traffic demands at the major and minor approaches (around 900 veh/hr), delays might
be underestimated due to the assumption of no initial queue. Average control delay (d) is estimated using the proposed function by
HCM 6th edition (2016) as shown in Eq. (10).
g 2

d=
(
0.5C 1− C ) (PF ) + 900T ⎡X −1 + (X −1)2 +
8kIX ⎤
1−⎡min(1,
g
X)C⎤ ⎢ cT ⎥
⎣ ⎦ (10)
⎣ ⎦
where C is cycle length (sec); g is effective green time (sec); X is degree of saturation; T is analysis period duration (h); PF is
progression adjustment factor (assumed as 1.0 for uniform arrivals); k is incremental delay factor (assumed as 0.5); I is upstream
filtering adjustment factor (assumed as 1.0), and c is lane group capacity (veh/h).

3.4. Pedestrian impedance

Pedestrians are main intersection road users. The conflicts between pedestrians and vehicles may significantly hinder the safety
and mobility levels of intersections. This deterioration in performance is varying from one control type to the other. Pedestrian
influence is incorporated for the adjustment of intersection capacity as shown in the following sub-sections.

3.4.1. Pedestrian consideration at RABs

The proposed methodology by HCM 6th edition (2016) for the estimation of RAB capacity considers pedestrian influence using
adjustment factor (fped) which is estimated under standard roundabout designs. Kang and Nakamura (2015) highlighted that the
proposed methodology by HCM 2010 underestimates the impact of pedestrians on the capacity of roundabouts. Thus, they proposed a

Fig. 5. Entry capacity considering pedestrian (Kang and Nakamura, 2015).

59
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

different approach to estimate roundabout entry capacity considering conflicts with pedestrians by considering gap acceptance
behavior. As shown in Fig. 5, circulating flow may encounter one of two conditions; flowing in case there are no crossing pedestrians
at the exit approaches or queuing due to the presence of crossing pedestrians at the exit approaches. Based on these conditions and
the gap acceptance theory, they proposed a procedure for the estimation of entry capacity. A short summary of their procedure and
the models used to estimate entry capacity, which are adopted in this study, are provided below.
Assuming ca is the capacity under flowing condition where there is no blocking of circulating flow movement because of the
presence of pedestrians at exit approaches, then the total entry capacity (ce) is given by Eq. (11).
ce = Pflowing ca=(1−Pqueuing ) ca (11)
where Pflowing is probability of flowing and Pqueuing is probability of queuing due to the presence of crossing pedestrians at exit
approaches. ca is given by Eq. (12). The first component describes maximum number of vehicles entering from storage space (the
space between yield line and the crosswalk at the entry approach) and the second component describes maximum number of vehicles
entering through the crossing pedestrians and the circulating vehicles without stopping in the storage space.
na 1
ca = f (qcir ) + g (qped , qcir )
na + 1 na + 1 (12)
where na is the maximum number of vehicles which can be stored in storage space; f(qcir) is entry capacity of vehicle from yield line
estimated by Eq. (13) and g(qped, qcir) is entry capacity of vehicle entering from stop line located before the crosswalk at the entry
approach and it is estimated by Eq. (14). For the purpose of this study, it is assumed that the available storage space at entry
approaches can accommodate one vehicle only (na = 1).

3600 q q t
f (qcir ) = ⎛1−τ cir ⎞ exp ⎡− cir ⎛tc− f −τ ⎞ ⎤
tf ⎝ 3600 ⎠ ⎢ 3600 ⎝
⎣ 2 ⎠⎦ ⎥ (13)
ped ped
g (qped , qcir ) = cmax (Poped cir cir cir
, s Po, B Po, Q )(Po, s Po, B Po, Q ) (14)
where tf is follow-up time of entry flow (sec); τ is minimum headway of circulating vehicles (sec); tc is critical gap (sec); qcir is
circulatory conflicting vehicles (veh/h); cmax is maximum entry flow rate under free flow condition (veh/h); Poped , s is probability of “no
, s is probability of “no single circulating vehicle”. Both Po, s and Po, s are the probability of headway t in the major
single pedestrian”; Pocir ped cir

flow larger than t0 under the condition that headway t is larger than minimum headway τ (P0, s = P(t > t0 /t > τ )); Here, t0 is the
intercept gap size with assumed continuous departure flow and estimated as t0 = tc−t f /2 (Siegloch, 1973).
, B is probabilty of “no bunching pedestrians”; Po, B is probabilty of “no bunching circulating vehicles”. They are estimated
Poped cir

dependent on average minimum headway τ and flow demand q (P0, B = 1−PB = 1−q × τ ).
Poped
, Q is probabilty of “no queuing pedestrians” and Po, Q is probability of “no queuing circulating vehicles”. They are estimated as a
cir

function of degree of saturation (P0, Q = 1−PQ = 1−X ). The adopted values for tf, τ , and tc are listed in Table 2. For further details on
the previous models, refer to Kang and Nakamura (2015).

3.4.2. Pedestrian consideration at TWSC intersections

For simplification, it is assumed that the influence of pedestrians at TWSC follows the same procedure of RAB by applying the gap
acceptance concept. The capacity of vehicle movement j (cj) without considering pedestrian effect is adjusted for the consideration of
crossing pedestrians following Eq. (15).
ped ped
cadj, j = cj ∗ (Poped
, s Po, B Po, Q ) (15)
where cadj,j is adjusted capacity for vehicle movement j due to the influence of pedestrians (veh/h); and cj is capacity of vehicle
movement j without considering pedestrian influence (veh/h). The definitions of other terms in Eq. (15) are similar to those in Eq.
(14). cj is estimated using HCM 6th edition (2016) procedure. Pedestrian gap parameters are adopted from Kang and Nakamura
(2015).

3.4.3. Pedestrian consideration at SIG intersections

Pedestrians have a considerable influence on the capacity of turning movements at signalized intersections. The influence of
pedestrian on the design of signal control is explicitly considered. The impact on vehicle saturation flow rate as well as minimum
pedestrian clearance time requirement for signal setting are incorporated. Since low to medium demand of pedestrians are assumed
in this study, intersection vehicle delay is used as the measure of effectiveness for the comparison between the three intersections
control types. The procedure proposed by HCM 6th edition (2016) for the estimation of saturation flow rate is used in this study to
reflect the influence of pedestrians. The manual proposed the adjustment factor (fp) to reflect the impact of pedestrians and it is
calculated by estimating average pedestrian occupancy at the conflict zone as shown in Eq. (16).
fp = 1−OCCr (16)
where OCCr is relevant conflict zone occupancy by pedestrians and it is a function of the service time provided for vehicular turning
movements. Furthermore, it is important to mention that pedestrian minimum green time requirement for the setting of signal timing
is considered as explained in Section 3.2 “Intersection Control Parameters”.

60
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

Table 5
Analysis scenarios.
Balanced Unbalanced
Equal turning ratios on major & minor streets Unequal turning ratios on major & minor streets

Major Street 0:100: 0 10: 80: 10 20: 60: 20 10: 80: 10

Minor Street 40: 20: 40
Pedestrian flow (ped/h) 0 0 100 200 0 0 100 200

3.5. Analysis scenarios

To compare the performance of various intersection control types, it is assumed that the major and minor streets have one-way
traffic volumes (entering to the intersection) varying from 0 to 1000 veh/h. Average intersection vehicle delay was estimated for all
combinations of vehicular traffic volumes from the major and minor streets in an increment of 10 veh/h. Each increment is analyzed
using four different combinations of turning volume percentages as shown in Table 5. For simplification, the percentage of LT traffic
is assumed to be equal to that of RT. As shown in Table 5, three balanced (equal percentages of turning traffic on the major and minor
streets) and one unbalanced (unequal percentages of turning traffic on the major and minor streets) combinations are analyzed.
Regarding the pedestrian flows, two levels are considered, 100 ped/h and 200 ped/h, as shown in Table 5. It is assumed that all
approaches of the intersection have equal pedestrian flows of either 100 ped/h or 200 ped/h. To limit the number of analysis
scenarios, pedestrian effect is considered in two vehicle turning percentage combinations (Table 5). In total, we have eight scenarios
for the analysis. It is important to note that the influence of bicycles and heavy vehicles are not considered.

4. Sensitivity analysis

4.1. Impact of turning traffic on intersection vehicle delay

Fig. 5 presents the estimated average intersection vehicle delay for the analysis scenarios when there are no pedestrians (Table 5).
Fig. 5(a) shows the estimated TWSC average intersection vehicle delay curves for different scenarios assuming a traffic flow of 800
veh/h in the major streets. It is clear that as the percentage of turning traffic (on major and minor streets) increases, average
intersection vehicle delay increases. This can be attributed to the increasing percentage of the lowest rank movements (through and
RT) on the minor streets which significantly increase the average intersection vehicle delay.
For RAB, the impact of turning traffic is found to be not significant when the traffic flow on the minor street is lower than 500
veh/h as shown in Fig. 6(b). However, with increasing traffic flow on the minor street, the delay increases exponentially which
indicates that RAB might not be the best control type in such case. As traffic flow increases at RAB approaches, the conflicts with
circulating flow also increase which reduce the entry capacity and increase intersection vehicle delay. Furthermore, Fig. 6(b) shows
that under high traffic flows, unbalanced turning percentage scenario results in significantly higher delays compared to other sce-
narios.
For SIG, the impact of turning traffic increases as the traffic flow increases at the intersection as shown in Fig. 6(c). The higher the
percentage of turning traffic, the larger the rate of increase in intersection vehicle delays. This is clear in the case of unbalanced
turning percentage scenario (purple dashed line in Fig. 6(c)) in which a turning ratio of 80% on the minor street is assumed. This
large turning ratio yields to long cycle lengths as well as large delays.

4.2. Impact of pedestrians on intersection vehicle delay

The impact of pedestrians is demonstrated in Fig. 7 in which the traffic flow on the major street is assumed as 800 veh/h and the
tuning percentages are assumed to follow the unbalanced scenario shown in Table 5 (Major = 10:80:10; Minor = 40:20:40). Presence
of pedestrians significantly affect average intersection vehicle delays for every control type, and particularly the performance of
TWSC is strongly affected by them as shown in Fig. 7(a). As pedestrian flow increases, the capacity of TWSC intersections significantly
deteriorates. Thus, it reaches oversaturation with a small increase of the minor street traffic flow.
In the case of RAB, similar tendencies of pedestrian flow impacts are observed. With increasing pedestrian flow, RAB delay
significantly increases. As shown in Fig. 7(b), intersection vehicle delay at RAB significantly increases when pedestrian flow changes
from 0 to 100 ped/h. This can be attributed to the increase in the conflicts between pedestrians and vehicular flows (entering and
exiting) which reduce the number of acceptable gaps and significantly increases intersection vehicle delay. Furthermore, knowing
that pedestrians have the right of way over entering and exiting vehicular flows, a significant reduction in the capacity is expected.
Whereas, SIG shows better absorption of pedestrian influence. At SIG intersections, only RT and LT traffics conflict with pedestrians
depending on the implemented phasing plan. Furthermore, the assumed intersection layout has crosswalk length of 9 m (Fig. 4(c))
which limits the impact of pedestrian minimum green time requirements. At large intersections with long crosswalks, pedestrian
minimum green time requirement becomes major influencing factor in the performance of the SIG intersection.

61
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

200 200
Balanced (Maj = Min) Balanced (Maj = Min)
180 180
Intersection delay (sec/veh)
0: 100: 0

Intersection delay (sec/veh)

0: 100: 0
160 10: 80: 10 160 10: 80: 10
140 20: 60: 20 140 20: 60: 20
120 Unbalanced Unbalanced
120
Maj = 10: 80: 10 Maj = 10: 80: 10
100 Min = 40: 20: 40 100 Min = 40: 20: 40
80 80
60 60 Adams model (no
40 40 turning traffic)
20 20
0 0

Minor street flow (veh/h) Minor street flow (veh/h)

a)TWSC b)RAB
200
Balanced (Maj = Min)
180 0: 100: 0
Intersection delay (sec/veh)

160 10: 80: 10

140 20: 60: 20
Unbalanced
120 Maj = 10: 80: 10
100 Min = 40: 20: 40
80
60
40
20
0

Minor street flow (veh/h)

c) SIG
Fig. 6. Impact of turning traffic on intersection vehicle delay (traffic flow at major street is 800 veh/h).

5. Comparison analysis

The comparisons of intersection vehicle delays among the three control types (TWSC, RAB, SIG) are presented in the following
two sections. The first section concentrates on the impact of turning traffic without considering pedestrian flows while the second
section considers the impact of pedestrians.

5.1. Intersection control performance without considering pedestrians

Fig. 8 demonstrates how the performance area of each control type is identified. The horizontal and vertical axes represent the
major and minor streets one-way (entering into the intersection) volumes, respectively. It presents the case of no turning traffic on
both major and minor streets (the first scenario in Table 5) and no pedestrians as well. The contour lines shown in Fig. 8(a) present
the delay of each control type (TWSC, RAB, SIG-2 and SIG-4). The blue solid contour lines are for SIG-2 (since there is no turning
traffic, SIG-4 is not feasible), while the green dashed contour lines are for RAB and the red dotted lines are for TWSC. By comparing
these contour lines, the control type that yields to the minimum intersection vehicle delay on each major and minor street flow
combination can be identified. Then, the area where each control type outperforms others (performance area) can be distinguished
and delineated as shown in Fig. 8(b). TWSC outperforms RAB and SIG when the traffic flow on the minor street is low. For SIG, the
estimated optimum cycle lengths when traffic flows are low are considerably shorter than the minimum cycle length requirements
based on the minimum vehicle and pedestrian green times. In this case, minimum cycle length requirement dominates which yields to
larger delays compared to other control types. Furthermore, the effect of RAB geometry in reducing vehicle speeds results in larger
delays compared to TWSC intersection. This explains why TWSC outperforms other control types at the low traffic flow on the minor
street.
RAB outperforms other control types when traffic flows on the major and minor streets are in the range of 200 and 500 veh/h as
shown in Fig. 8(b). As the traffic flow increases over 500veh/h, available gaps between vehicles on circulatory roadway significantly
decrease which reduce the entry capacity and increases intersection vehicle delay. Due to that, RAB performance area deteriorates

62
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

Fig. 7. Impact of pedestrian flow on intersection vehicle delay for turning percentage scenario (10:80:10) assuming the flow at major street is 800
veh/h.

Fig. 8. Control type comparison for the no turning traffic scenarios (0:100:0) at the major and minor streets.

while the performance of SIG dominates.

Fig. 9 demonstrates the impact of turning traffic on the performance areas of intersection control types. The performance area of
SIG is separately indicated for SIG-2 and SIG-4. In general, Fig. 9 shows that the performance areas vary with the turning traffic

63
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

Fig. 9. Performance area of different control types without considering pedestrian impact.

percentages. In all turning percentage scenarios, TWSC outperforms other control types when the traffic flow on the minor street is
low (less than 200 veh/h). This is expected due to assumption of the absolute right of way of major street vehicles. However, as the
traffic flow on the minor street increases, the performance of TWSC declines.
By comparing Fig. 8(b) with Fig. 9(a) and (b), it is clear that turning traffic negatively affects the performance area of RAB.
However, there is no significant change when turning traffic ratio changes from 20% (Fig. 9(a)) to 40% (Fig. 9(b)). This tendency
completely changes in case of unbalanced turning percentage scenario (Fig. 9(c)) where the minor street has a total turning per-
centage of 80%. In this case, the performance areas of RAB and TWSC significantly increases. This is attributed to a significant
deterioration in the performance of SIG at high turning traffic percentages. Regardless of the phasing plan, high percentages of RT
traffic on the minor street significantly increase the delays at SIG intersections. Furthermore, at high RT traffic percentage at the
minor approach, SIG-4 performance area significantly increases over SIG-2, since the four-phase plan with exclusive right turn
phases, becomes more appropriate to accommodate the high RT demand.
To investigate the reliability of the borderlines between intersection control types, the area that corresponds to the absolute
difference of 1.0 sec delay (gray shaded area) is presented in Fig. 9 and it is named in this study as gray area to facilitate the
discussion. In the case of balanced turning percentage scenarios (Fig. 9(a) and (b)), the gray area between RAB and SIG-2 is wide
which indicates that both control types performs closely in this region. However, in the case of unbalanced turning percentage
scenario where the minor street has high turning percentage (80%), the gray area between RAB and SIG-2 is narrow which indicates
that there are significant changes in the performance of each control type around that boundary.
Fig. 10 compares the performance areas referred to in the Japan RAB Manual (2016) with those estimated in this study, without
the consideration of turning traffic and pedestrian flows. It is clear that the performance area of TWSC is very similar. However, the
estimated performance area of RAB is significantly smaller than that indicated in the Japan RAB Manual (2016). This is mainly
attributed to the different procedures for the estimation of RAB delay. The referred figure in Japan RAB Manual (2016) utilized
Adams delay model (Troutbeck, 1986) to estimate RAB delay. Several studies stated that Adams model significantly underestimates

64
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

Fig. 10. Comparisons between the proposed selection chart and Japan RAB Manual (2016) without considering turning traffic and pedestrian flows.

intersection vehicle delay at medium and high traffic flows (Troutbeck, 1986; Troutbeck and Brilon, 2008). To validate that, in-
tersection vehicle delays are estimated using Adams model as shown in Fig. 6(b). It confirms that Adams model significantly un-
derestimates the delay.
By comparing the results of this study (Figs. 8(b) and 9(b)) with those of Bumjon et al (2017) presented in Fig. 3, it is clear that
there are significant differences. Bumjon et al (2017) concluded that when signal control is optimized, SIG and RAB outperform
TWSC in all cases which is the reason why TWSC does not appear in Fig. 3. This contradicts with previous studies (Prakash et al.,
2014; Japan RAB Manual, 2016) as well as the results of our study. This can be attributed to the underestimated RAB delay in Bumjon
et al (2017) caused by the missing geometric delay component in RAB delay estimation methodology. However, the lines delineating
the performance areas of RAB and SIG in both studies are close to each other. Furthermore, it is important to mention that our study
indicates significant impact of turning traffic on the performance areas of intersection control types which is in accordance with
Prakash et al. (2014). This contradicts with the outcomes of Bumjon et al (2017) in which insignificant impact of the percentage of
left turning traffic on the performance areas of intersection control types (Fig. 3) is concluded.

5.2. Intersection control performance with considering pedestrian

Fig. 11 shows the performance areas of various control types considering two pedestrian flows; 100 ped/h and 200 ped/h per
intersection approach. Fig. 11(a) and (b) are for the balanced turning percentage scenario (10:80:10) while Fig. 11(c) and (d) are for
the unbalanced turning percentage scenario (Major = 10:80:10; Minor = 40:20:40). It is observed that pedestrians have notable
influences on the performance areas of the control types. For the case of balanced turning percentage scenario by comparing Fig. 9(a)
with Fig. 11(a) and (b), the performance area of TWSC significantly shrinks as pedestrian flow increases which is also similar for RAB.
The presence of pedestrians block vehicles from entering the TWSC or RAB, thus they have to wait until the acceptable gaps between
consecutive pedestrians appear. This phenomenon deteriorates capacity of TWSC and RAB, and significantly increases the vehicle
control delays. However, only turning traffic conflicts with pedestrians at SIGs, which leads to fewer conflicts compared to other
control types. With increasing pedestrian flow from 100 ped/h to 200 ped/h, the performance area of TWSC significantly decreases
while the boundary between RAB and SIG slightly shifts.
The same tendency can be found in the unbalanced turning percentage scenario by comparing Fig. 10(c) with Fig. 11(c) and (d)).
The performance areas of TWSC and RAB significantly shrink as pedestrian flow increases, while the area of SIG expands. However,
the borderline between RAB and SIG does not change significantly when pedestrian flow increases from 100 to 200 ped/h. Mean-
while, the performance area of SIG-2 expands over that of SIG-4 at 100 ped/h demand level compared to the case of no pedestrians
(Fig. 9(c)). This can be explained that SIG-2 can accommodate low pedestrian flow without causing large increase in intersection
vehicle delay. Whereas in SIG-4, the minimum pedestrian green time affects the green time of the through-left turn phase which leads
to higher delays.

6. Conclusion and future work

This study assessed and compared the operational performance of three intersection control types: TWSC, RAB and SIG. Average
intersection vehicle delay was used as the measure of effectiveness for the operational performance evaluation. Various scenarios of
turning traffic percentages were analyzed. Additionally, the impact of pedestrian flows on intersection performance was also eval-
uated by assuming two pedestrian flow levels (100 and 200 ped/h per approach). In the signal timing optimization of SIG inter-
sections, two-phase and four-phase plans were considered. Finally, charts for the selection of intersection control types at various
turning traffic percentages and pedestrian flows were developed.
It is concluded that turning traffic and pedestrian flows significantly affect the performance areas of intersection control types. As
the percentage of turning traffic increases, the area where RAB outperforms other control types shrinks while SIG performance area

65
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

Fig. 11. Performance area of different control types considering pedestrian impact.

grows. Whereas, in case of high turning traffic on minor streets only, the performance area of RAB grows due to the significant
deterioration in the performance of SIG. Furthermore, this study showed that pedestrians negatively affect the performance of TWSC
and RAB where SIG becomes advantageous since it can separate the rights of way of pedestrians and some vehicle movements in
different phasings, which can limit the conflicts between them.
In general, the outcomes of this study can support engineers at the planning stage to select a proper control type for intersection.
However, this study focused on isolated intersections without considering the potential impacts of neighboring intersections, which
are important factors especially in urban areas. Furthermore, it is important to investigate the impact of intersection size on the
relative performance of various control types. The calibration of model parameters using real world data, especially gap parameters
for pedestrian impact, is necessary. Furthermore, the roundabout geometric delay model adopted in this paper need to be validated in
the future. Additionally, the estimation of person delay using different vehicle occupancies rates and pedestrian delay is an important
aspect that will be investigated. Moreover, the utilization of different measures of effectiveness such as intersection capacity is
necessary to provide a holistic assessment of the relative performance of the intersection control types.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.tra.2018.08.039.

References

Andrzej, T., Jing, L., Nagui, R., 2001. Traffic Flow Theory at Signalized Intersection. North Carolina University, Raleigh, NC.

66
A.N. Arshi et al. Transportation Research Part A 118 (2018) 52–67

Bumjon, B., Brandon, C., Lee, D., 2017. Gray areas in isolated intersection control-type selection: complementary decision-support tool. J. Transport. Eng., Part A: Syst.
143 (11).
Japan Society of Traffic Engineers, 2006. Manual on Traffic Signal Control, Revised Edition (MTSCRE) (in Japanese).
Japan Society of Traffic Engineers, 2016. Japan Roundabout Manual, Tokyo (in Japanese).
Kang, N., Nakamura, H., 2015. Estimation of roundabout entry capacity that consider conflict with pedestrians. Transportation research record. J. Transport. Res.
Board No 2517. Washington, D.C. 61–70.
Kyte, M., Joseph, F.M., Tian, Z., Kent, L., Kenneth, V., 1997. Determining Intersection Traffic Control Using the 1994 Highway Capacity Manual. Institute of
Transportation Engineering, pp. 22–26.
Lee, D., Jan-Mou, L., Thomas, U., 2008. Control-type selection at isolated intersections based on control delay under various demand levels. Transportation research
record. J. Transport. Res. Board, No 2071. Washington D.C 109–116.
Manual of Uniform Traffic Control Device for Streets and Highways (MUTCD), 1988. Department of Transportation. Federal Highway Administration. Washington D.C.
McDonald, M., Housell, N., Kimber, 1984. Geometric Delay at Non-Signalized Intersections. Springer Publication, Berlin.
Prakash, R., Shahin, A., Fasihullah, S., 2014. Evaluating operational performance of intersection using SIDRA. Open Transport. J. 8, 51–62.
Siegloch, W., 1973.: Die Leistungsermittlung an Knotenpunkten Ohne Lichtsignalsteuerung. Schriftenreihe Strassenbau und Strassenverkehrstechnik, Vol. 154 (in
German).
Transportation Research Board, 2000. Highway Capacity Manual, Washington D.C.
Transportation Research Board, 2016. Highway Capacity Manual 6th edition, Washington D.C.
Troutbeck, R.J., Brilon, W., 2008. Unsignalized Intersection Theory. Springer, Berlin.
Troutbeck, R.J., 1986. Average delay at an unsignalized intersection with two major streams each having a dichotomized headway distribution. Transport. Sci. 20 (4),
272–286.
Webster, F.V., 1966. Traffic Signal. Road Research Technical Paper No. 56. Road Research Laboratory. Her Majesty’s Stationary Office, London.
Yoshioka, K., Nakamura, H., Mabuchi, T., 2008. A discussion on optimum intersection control type under low traffic based on delay and environmental impact. In:
Proceedings of the 28th Meeting of Japan Society of Traffic Engineers, 37–40 (in Japanese).

67