zhao2014

Transportmetrica B: Transport Dynamics, 2014
Vol. 2, No. 2, 130–150, http://dx.doi.org/10.1080/21680566.2014.908751
Integrated design and operation of urban arterials with reversible lanes

Jing Zhaoa , Wanjing Maa∗ , Yue Liub and Xiaoguang Yanga
a Key Laboratory of Road and Traffic Engineering of the Ministry of Education, Tongji University,
4800 Cao’an Road, Shanghai, People’s Republic of China; b Department of Civil Engineering and
Mechanics, University of Wisconsin at Milwaukee, P.O. Box 784, Milwaukee, WI, USA
(Received 9 October 2013; accepted 23 March 2014)
Reversible lane operation, an effective strategy to relieve traffic congestion in urban arterials,
must interact with other conventional traffic management and control components in the arterial,
including left-turn restriction, lane channelisation, and signal timings. This paper develops a
lane-based optimisation model to guide the integrated setting of reversible lanes and other traffic
management measures in an arterial to maximise its operational performance. The optimisa-
tion problem is formulated as a multi-objective mix-integer non-linear programming model,
which is then transformed into a single-objective mix-integer linear programming formula-
tion for an optimal solution. Results from extensive numerical analyses have demonstrated the
effectiveness of the proposed model.
Keywords: reversible lanes; urban arterials; lane markings; capacity; traffic management
and control
1. Introduction
High traffic demand coupled with unbalanced directional flows on roadways exacerbates the
perennial problem of congestion. Reversible lanes, deemed by the Institute of Transportation
Engineers (ITE), as one of the most effective methods to increase peak-hour capacities of roadways
(Wolshon and Lambert 2006) could essentially accommodate the imbalanced traffic flows between
two driving directions on a congested roadway section (Tuydes 2005).
Lane reversal has been used for several decades and much has been investigated regarding its
effectiveness, feasibility and safety. One of the earliest referenced usages of reversible roadways
was in Los Angeles in 1928, with a convertible lane variant known as off-centre lane movement
(Dorsey 1948). Over the past decades, many different forms of reversible roadways have been
used throughout the world to address a variety of needs (Wolshon and Lambert 2004). Three
major applications have been developed to accommodate the demand associated with frequent
and predictable imbalanced peak-period travel times (Upchurch 1975; Agent and Clark 1980),
special events such as those associated with large sporting events, concerts, and festivals (Gillis
1990; Lambert and Wolshon 2002; Wolshon 2002), and emergency conditions such as evacuation
(Ford, Henk, and Barricklow 2000; PBS&J 2000; Wolshon 2001; Urbina and Wolshon 2003).
The principle of reversible roadways is to configure the lanes of a roadway to match available
capacity with traffic demand. The common idea among these designs is that they only change
the direction of the reversible lanes during daily peak periods, before and after large events, or
during emergency evacuations. For convenience of application, many professional transportation
∗ Corresponding author. Email: mawanjing@tongji.edu.cn
© 2014 Hong Kong Society for Transportation Studies Limited

Transportmetrica B: Transport Dynamics 131
organisations have also developed guidelines for operating reversible lanes, such as AASHTO
(2011), ITE (2009), JSCE (1989), and FGSV (2006).
Several optimisation models have been proposed to produce the best operational strategy for an
arterial with reversible lanes. Most of them formulate the optimal decision on the reversible lanes
as a network design problem and optimise a target system performance measure such as the total
system travel cost (Yang and Bell 1998). Tuydes and Ziliaskopoulos (2004) proposed a system-
optimal dynamic traffic assignment-based optimal capacity reversibility model (SODTA-CR)
to capture spatiotemporal changes in disaster conditions. Tuydes (2005) proposed three exten-
sions of SODTA-CR: lane-based capacity reversibility (SODTA-LCR), total-or-nothing capacity
reversibility (SODTA-TCR), and budgeted capacity reversibility (SODTA-BCR). These were
single-level models because the drivers and the evacuation manager share the same objective in
minimising total system travel time. For operational design, however, user equilibrium is consid-
ered as a more suitable principle to capture travellers’ route choice behaviour. Therefore, bi-level
programming models were proposed to capture different perspectives between authorities and
travellers. In a bi-level programming model developed by Zhang and Gao (2007), the upper level
model was to minimise the total system cost and the lower level was a user equilibrium assign-
ment. The model was solved using a particle swarm optimisation technique. Wu et al. (2009)
introduced flow entropy in the upper level objective function to obtain symmetrical flows. The
upper level of Wu’s model was to minimise the total system cost and flow entropy and the lower
level was a stochastic user equilibrium assignment with an advanced traveller information sys-
tem. This model was solved using a chaotic optimisation algorithm. Xie, Lin, and Waller (2010)
discussed a dynamic evacuation network optimisation problem that incorporated lane-reversal
and crossing-elimination strategies. A bi-level network optimisation model was formulated in
which the upper level was aimed at optimising the network evacuation performance subject to
the lane-reversal and crossing-elimination constraints and the lower level was conveyed a cell-
transmission-based dynamic traffic assignment problem. An integrated Lagrangian relaxation
and tabu search method were designed for approximating the optimal problem solutions (Xie and
Turnquist 2011). Karoonsoontawong and Lin (2011) allowed time-varying reversibility with dif-
ferent reversibility durations for various candidate link pairs in the bi-level programming model,
such that the optimal starting times and the optimal reversibility durations for candidate link pairs
could be determined for peak-period traffic management on a daily basis. A genetic algorithm
with a cell-transmission-based user-optimal dynamic traffic assignment was adopted to solve the
problem.
In summary, most existing literatures on reversible lane operation focus on optimising the
number and the selection of lanes to be reversed. As a special type of traffic management strat-
egy, however, implementation of reversible lanes must interact with other conventional traffic
management and control strategies, such as left-turn restriction, lane channelisation, and signal
timing. Neglecting such interactions may result in non-optimal design results and unsatisfactory
operational performance in the arterial. In previous studies (Wong and Wong 2002; Wolshon and
Lambert 2004), integration of these traffic management strategies was usually conducted sequen-
tially, not simultaneously. Direction of reversible lanes can first be determined using discrete
network design formulations, in which the effect of signalised intersections is simplified. Then,
intersection signal timings are designed based on the turning restriction and lane channelisation
plans that have been determined. The above step-by-step design procedure also assumes relative
independence among different traffic management strategies and therefore cannot reflect their
interactions under different traffic flow patterns.
To remedy the above deficiency, researchers have developed the integrated design concept and
applied it to isolated intersections. Several integrated models (Lam, Poon, and Mung 1997; Wong
and Wong 2003a, 2003b; Wong and Heydecker 2011; Zhao et al. 2013a) have been established
132 J. Zhao et al.
to combine the design of lane markings and signal timings for isolated signalised intersections.
It was shown that substantial improvement in the intersection performance could be achieved by
using the integrated model.
Despite the promising property of these integrated design models, only limited studies have
been done regarding the integrated design and operation of an urban arterial with reversible lane
settings, turning restrictions, lane channelisation, and signal timing plans optimised in a unified
framework. This can best demonstrate its effectiveness when there exist significant directional
flow patterns in the arterial and the improvement of its operational performance is constrained
by the available capacity at certain bottleneck intersections. In response to the above research
needs, this paper develops a lane-based optimisation model to simultaneously determine the lane
assignments at roadway segments and intersections, left-turn restrictions, and signal timings in a
target urban arterial.
The remainder of this paper is organised as follows. In Section 2, the model configuration
and assumptions are described. The integrated optimisation model is proposed in Section 3.
The performance of the integrated model is evaluated through numerical analysis in Section 4.
Conclusions and recommendations are given at the end of the paper.
2. Model configuration and assumptions

This study focuses on an arterial that includes both segments with reversible lanes and intersections
along the arterial (Figure 1). The median lanes in the internal segments are used as reversible lanes.
The lane markings of all the approach lanes are also flexible and can be changed dynamically.
The objective of this research is to develop an integrated model that can simultaneously optimise
the number of lanes for each direction of the segment, the turning restriction strategies, the lane
markings for each approach lane, and the signal timing for each intersection.
To ensure that the proposed model is tractable and reflects real-world constraints, this study
has employed the following assumptions:
Figure 1. Configuration of the integrated model.

(1) Traffic demand distribution in the target arterial is assumed to be given.

(2) If a left turn at an intersection approach on the arterial is prohibited, the left-turn movement
will travel through the intersection and make a U-turn at the nearest downstream intersection
at which a left turn is permitted.
(3) The assigned direction of the reversible lanes within a road segment should be continuous.
3. Integrated optimisation model

3.1. Notations
To facilitate the model presentation, notations and parameters used hereafter are listed in Table 1.
3.2. Objective function

Existing literature (Wolshon and Lambert 2006; Yu and Prevedouros 2013; Zhao et al. 2013b)
has shown that the setting of reversible lanes, dynamic lane assignment, and left-turn restrictions
can increase the capacity of a target arterial. These may also, however, cause potential driver
confusion issues. Traffic management strategies such as lane reversion and turning restriction
should not be considered, therefore, until a conventional design can no longer accommodate the
traffic demand. Four control objectives, with different priority levels, are considered, therefore,
to accommodate different perspectives of authorities and road users.
4 These
3 are: (a) maximising
the number of permitted movements at the intersections, Nr=1 r
w=1 δriw , (b) minimising
Nr −1
i=1
the number of direction adjustments in the segment lanes, r=1 γr(r+1) , (c) maximising the
reserve capacity
ofr the arterial, μ, and (d) maximising the summation of reserve capacities for all
intersections, Nr=1 μr . The first two objectives reflect the perspective of roadway users. The last
two are designed to improve the overall system performance set by the responsible agencies. In
summary, the overall control objective function is stated as:

Nr
4
3
N r −1
Nr
max P1 δriw − P2 γr(r+1) + P3 μ + P4 μr , (1)
r=1 i=1 w=1 r=1 r=1
where P1 , P2 , P3 , P4 are priority coefficients for the four control objectives, P1 P2 P3 P4 ;

δriw is the permission of movement w in arm i at intersection r (1 − Yes, 0 − No); Nr is the number
of signalised intersections in the target arterial; γr(r+1) is the adjustment of direction of reversible
lanes in segment (r, r + 1), if adjusted γr(r+1) = 1, if not adjusted γr(r+1) = 0; μ is the common
flow multiplier for the target arterial; and μr is the common flow multiplier for intersection r.
3.3. Constraints
3.3.1. Assigned flow constraints
The assigned flows on different lanes, qriwk , should obey the following set of constraints.
(1) The sum of the assigned flows on the different lanes should be equal to the increased
demand for that movement.
nriaf +nriac

μr Qriw = qriwk , ∀ r ∈ N ; i ∈ Ar ; w ∈ Ti , (2)
k=1

where Qriw is the adjusted demand flow of movement w in arm i at intersection r considering
left-turn prohibition (veh h−1 ); nriaf is the number of fixed approach lanes in arm i at
134 J. Zhao et al.
Table 1. Notations.
Network representation
N Set of intersections
r∈N Index of signalised intersections in the target arterial (intersection r located left of
intersection r+1)
S Set of segments
(r, r + 1) ∈ S The road segment between intersection r and r + 1
Ar Set of arms in intersection r
i ∈ Ar Index of intersection arms, i = 1 for east arm, i = 2 for south arm, i = 3 for west arm,
and i = 4 for north arm, and the east and west arms are the arterial arms, the south and
north arms are the cross-street arms, shown in Figure 1
Ti Set of turning movements in arm i
w ∈ Ti Index of turning movements in arm i, w = 1 for U-turn and left-turn, w = 2 for through
movement, and w = 3 for right-turn
k Index of lanes, numbering from the left-most lane, as in Figure 1
Parameters
Nr Number of signalised intersections in the target arterial
af
nri Number of fixed approach lanes in arm i at intersection r
ef
nri Number of fixed exit lanes in arm i at intersection r
ac
nri Number of reversible lanes used for approach lanes in arm i at intersection r
ec
nri Number of reversible lanes used for exit lanes in arm i at intersection r
nr(r+1),(r+1)r Number of reversible lanes in segment (r, r + 1)
e
nriw Number of lanes on the exit that receives turning w on arm i
δriw Permission of movement w in arm i at intersection r (1 – Yes, 0 – No)
γr(r+1) Adjustment of direction of reversible lanes in segment (r, r + 1), if adjusted γr(r+1) = 1,
if not adjusted γr(r+1) = 0
μ Common flow multiplier for the target arterial
μr Common flow multiplier for intersection r
qriwk Assigned flow of movement w on lane k in arm i at intersection r (veh h−1 )
Qod Demand flow from demand origin o to destination d (veh h−1 )
Qriw Demand flow of movement w in arm i at intersection r (veh h−1 )

Qriw Adjusted demand flow of movement w in arm i at intersection r considering left-turn
prohibition (veh h−1 )
od
αriw A binary indicator showing if the OD pair (o, d) makes turning w on arm i at intersection
r without considering left-turn prohibition (1 – Yes, 0 – No)
srik Saturation flow rate of lane k in arm i at intersection r (veh h−1 )
dsmax Maximum acceptable degree of saturation
yrik Flow ratio of lane k in arm i at intersection r
Cmin , Cmax Minimum and maximum cycle length (s)
Ir(iw,i w ) Clearance time for a pair of conflicting traffic movements (s), w ∈ Ti , w ∈ Ti
M A large positive constant
Decision variables
xriwk A binary variable indicating the permission of movement w on lane k in arm i at
intersection r (1 – Yes, 0 – No)
nr(r+1) Number of reversible lanes used for the direction from intersection r to intersection r + 1
n(r+1)r Number of reversible lanes used for the direction from intersection r + 1 to intersection r
ξr Reciprocal of signal cycle length at intersection r (s−1 )
griw Start of green for movement w in arm i at intersection r
λriw Green time ratio for movement w in arm i at intersection r
Grik Start of green on lane k in arm i at intersection r
rik Green time ratio on lane k in arm i at intersection r
intersection r; and nriac is the number of reversible lanes used for approach lanes in arm i
at intersection r.
(2) The reverse capacity of the target arterial is determined by the critical intersection with the
lowest reserve capacity. By adopting the commonly used assumption that the proportions
of the traffic demand remain constant, maximising the reserve capacity is equivalent to
maximising the common flow multiplier μ (Gallivan and Heydecker 1988; Wong and Wong
2003a, 2003b; Wong and Heydecker 2011). However, there are several intersections in the
target arterial, and each intersection has its own reserve capacity, μr . Therefore, the reserve
capacity of the target arterial determined by the intersection with lowest μr . Moreover,
since the initial traffic demand must be accommodated, the common flow multiplier should
be no smaller than 1
μr ≥ μ ≥ 1, ∀r ∈ N. (3)
(3) Right-of-way constraint: If xriwk = 0, the movement w in arm i does not have the right-of-
way on lane k, so the assigned lane flow will be 0
Mxriwk ≥ qriwk ≥ 0, ∀ r ∈ N ; i ∈ Ar ; w ∈ Ti ; k ∈ {1, . . . , nriaf + nriac }, (4)
where xriwk is a binary variable indicating the permission of movement w on lane k in arm
i at intersection r (1 − Yes, 0 − No); and M is a large positive constant.
(4) Determining the original traffic demand without consideration of left-turn prohibition:
given the set of demand origins and destinations, Qod , as an exogenous input, the demand
flow Qriw of movement w in arm i intersection r can be obtained according to the following
equation:

Qriw = od
αriw Qod , ∀ r ∈ N ; i ∈ Ar ; w ∈ Ti , (5)
(o,d)
od
where αriw is a binary indicator showing if the OD pair (o, d) makes turning w on arm i at
intersection r without considering left-turn prohibition (1 − Yes, 0 − No).

(5) Determining the adjusted traffic demand, Qriw , considering left-turn prohibition: if a left
turn from the arterial in arm i intersection r is prohibited, the traffic demand should be
transferred to its detour path. It is assumed that all the vehicles of the prohibited movement
must first travel through the intersection, then execute a U-turn at the nearest downstream
intersection at which a left turn is permitted. Constraints (6) and (7) represent the adjusted
left-turn flow from the arterial arms. Constraints (8) and (9) represent the adjusted through
movement flow from the arterial arms. Constraints (10) and (11) represent the adjusted
right turn flows from the arterial arms. Finally, Constraint (12) represents the adjusted flow
from cross-street arms.

Nr t

Qr11 = δr11 Qr11 + (1 − δs11 ) Qt11 , ∀ r ∈ N , (6)
t=r+1 s=r+1

r−1
r−1

Qr31 = δr31 Qr31 + (1 − δs31 ) Qt31 , ∀r ∈ N, (7)
t=1 s=t

Nr
t
r−1
r

Qr12 = Qr12 + (1 − δs11 )Qt11 + (1 − δs31 )Qt31 , ∀r ∈ N, (8)
t=r s=r t=1 s=t

Nr
t
r
r

Qr32 = Qr32 + (1 − δs11 )Qt11 + (1 − δs31 )Qt31 , ∀r ∈ N, (9)
t=r+1 s=r t=1 s=t

Qr13 = Qr13 + (1 − δr31 )Qr31 , ∀r ∈ N, (10)
136 J. Zhao et al.

Qr33 = Qr33 + (1 − δr11 )Qr11 , ∀r ∈ N, (11)

Qriw = Qriw , ∀ r ∈ N ; i = 2, 4; w ∈ Ti . (12)
3.3.2. Lane assignment constraints

The assigned direction of the reversible lanes and the permission of movement w on lane k in arm
i at intersection r should obey the following set of constraints.
(1) Total number of reversible lanes: The total number of reversible lanes for a segment,
nr(r+1),(r+1)r , which is given as an exogenous input, should be equal to the sum of the
number of reversible lanes assigned to the two directions
nr(r+1),(r+1)r = nr(r+1) + n(r+1)r , ∀r ∈ N, (13)
where nr(r+1) is the number of reversible lanes used for the direction from intersection r
to intersection r + 1; and n(r+1)r is the number of reversible lanes used for the direction
from intersection r + 1 to intersection r.
(2) Continuous setting of the assigned direction for the reversible lanes: Constraints (14) and
(15) indicate that the assigned directions of the reversible lanes between two adjacent
intersections must be the same. Constraints (16)–(18) indicate that if the directions of
reversible lanes in two separate segments are changed, the directions of the reversible
lanes in the internal segments must also be changed
ac
nr1 = n(r+1)3
ec
= n(r+1)r , ∀ r ∈ {1, . . . , Nr − 1}, (14)
ac
n(r+1)3 = ec
nr1 = nr(r+1) , ∀ r ∈ {1, . . . , Nr − 1}, (15)
−M(2 − γr(r+1) − γs(s+1) ) ≤ (nr(r+1) − n(r+1)r ) − (nt(t+1) − n(t+1)t )
≤ M(2 − γr(r+1) − γs(s+1) ),
∀r, s, t ∈ {1, . . . , Nr − 1}; r ≤ t ≤ s, (16)
Mγr(r+1) ≥ nr(r+1) − n(r+1)r , ∀ r ∈ {1, . . . , Nr − 1}, (17)
Mγr(r+1) ≥ n(r+1)r − nr(r+1) , ∀ r ∈ {1, . . . , Nr − 1}, (18)
where nriec is the number of reversible lanes used for exit lanes in arm i at intersection r.
(3) Minimum number of permitted movements on traffic lanes: Each lane should carry at least
one movement, which can be specified as follows:

3
xriwk ≥ 1, ∀ r ∈ N ; i ∈ Ar ; k ∈ {1, . . . , nriaf + nriac }. (19)
w=1
(4) Minimum number of exit lanes: Due to safe and operational considerations, the number of
e
lanes at a movement’s corresponding exit arm, nriw , should always be at least as many as
the total number of lanes assigned to permit such a movement
nriaf +nriac

e
nriw ≥ xriwk , ∀ r ∈ N ; i ∈ Ar ; w ∈ Ti . (20)
k=1
(5) Conflict avoidance within an arm: For any two adjacent traffic lanes, k (left-hand) and
k + 1 (right-hand) from arm i, if the traffic movement of turn w is permitted on lane k + 1,
then traffic movements of all the other turns, w + 1, . . . , 3, should be prohibited on lane k
to eliminate potential internal cross-conflicts within the arm
1 − xriw(k+1) ≥ xriw k , ∀ r ∈ N ; i ∈ Ar ; w ∈ {1, 2}; w

∈ {w + 1, 3}; k ∈ {1, . . . , nriaf + nriac − 1}. (21)
(6) Number of lanes assigned to prohibited movements should be equal to 0
δriw ≥ xriwk , ∀ r ∈ N ; i ∈ Ar ; w ∈ Ti ; k ∈ {1, . . . , nriaf + nriac }. (22)
3.3.3. Signal timing constraints

The signal timing factors, including cycle length, phase plan, start of green and green time ratio,
should obey the following set of constraints.
(1) Cycle length: Let the minimum and maximum cycle lengths of the intersection be Cmin
and Cmax . Instead of defining the cycle length directly as a control variable, its reciprocal,
ξr , is used to preserve linearity in the mathematical formulation
1 1
≥ ξr ≥ . (23)
Cmin Cmax
(2) Start of green, griw : Since the signal timing at the intersection is cyclical, the start of
green can be picked arbitrarily along the time axis as long as it satisfies the other relevant
constraints in the formulation. However, for convenience, all of the starts of green variables
are confined within a signal cycle between 0 and 1
1 ≥ griw ≥ 0, ∀r ∈ N ; i ∈ Ar ; w ∈ Ti . (24)
(3) Green time ratio, λriw : The green time ratio of a movement should not be greater than 1.
Moreover, if a left turn from the arterial is prohibited, its green time ratio should be equal
to 0
1 ≥ λriw ≥ 0, ∀ r ∈ N ; i ∈ Ar ; w ∈ Ti , (25)
Mδriw ≥ λriw ≥ −Mδriw , ∀ r ∈ N ; i ∈ Ar ; w ∈ Ti . (26)
(4) Lane signal timing: If a lane is shared by more than one movement, these movements must
receive identical signal indications to avoid ambiguity. Considering a lane k from arm i,
if a movement w is permitted on this lane, then the following two constraint sets can be
established to fulfil the above condition:
M(1 − xriwk ) ≥ Grik − griw ≥ −M(1 − xriwk ),

∀ r ∈ N ; i ∈ Ar ; w ∈ Ti ; k ∈ {1, . . . , nriaf + nriac }, (27)
M(1 − xriwk ) ≥ rik − λriw ≥ −M(1 − xriwk ),
∀ r ∈ N ; i ∈ Ar ; w ∈ Ti ; k ∈ {1, . . . , nriaf + nriac }, (28)
where Grik is the start of green on lane k in arm i at intersection r; and rik is the green
time ratio on lane k in arm i at intersection r.
(5) Phase plan: Local practice is an important determinant in the selection of a phase plan
and phasing throughout an area should be relatively uniform (TRB 2000). In this model,
138 J. Zhao et al.
a dual-ring concurrent phasing scheme with assigned movements is adopted. This can be
specified as follows:
gr11 = gr31 = 0, ∀r ∈ N, (29)
gr32 = gr11 + λr11 + Ir(11,32) δr11 , ∀r ∈ N, (30)
gr12 = gr31 + λr31 + Ir(31,12) δr31 , ∀r ∈ N, (31)
gr32 + λr32 = gr12 + λr12 , ∀r ∈ N, (32)
gr21 = gr32 + λr32 + Ir(32,21) , ∀r ∈ N, (33)
gr41 = gr12 + λr12 + Ir(12,41) , ∀r ∈ N, (34)
gr42 = gr21 + λr21 + Ir(21,42) , ∀r ∈ N, (35)
gr22 = gr41 + λr41 + Ir(41,22) , ∀r ∈ N, (36)
gr42 + λr42 + Ir(42,11) = gr22 + λr22 + Ir(22,31) = 1, ∀r ∈ N . (37)
3.3.4. Flow ratio constraints

The flow ratio of a lane, yrik , is the ratio of the flow rate to the saturation flow rate, srik , that can
be calculated by Equation (38)
3
qriwk
yrik = w=1 , ∀ r ∈ N ; i ∈ Ar ; w ∈ Ti ; k ∈ {1, . . . , nriaf + nriac }. (38)
srik
For any approach, it is required that the degrees of saturation must be identical for a pair of
adjacent lanes having a common lane marking. Moreover, the signal settings on this pair of
adjacent lanes, as set in Constraints (27) and (28), must be equal. Therefore, to ensure identical
degrees of saturation, it is necessary that the flow ratios must be identical for a pair of adjacent
lanes having a common lane marking
M(2 − xriwk − xriw(k+1) ) ≥yri(k+1) − yrik ≥ −M(2 − xriwk − xriw(k+1) ),
(39)
∀ r ∈ N ; i ∈ Ar ; w ∈ Ti ; k ∈ {1, . . . , nriaf + nriac − 1}.
3.3.5. Degree of saturation constraints

In order to ensure that an intersection performs reasonably well, the degree of saturation of every
traffic lane should be no greater than the maximum acceptable limit, dsmax
dsmax rik ≥ yrik , ∀ r ∈ N ; i ∈ Ar ; k ∈ {1, . . . , nriaf + nriac }. (40)
3.4. Solution algorithm

The integrated optimisation model is a multi-objective mixed-integer non-linear programming
problem with the objective function of Equation (1) and Constraints (2)–(40). Since the four
goals are divided into four priority levels, it is a pre-emptive goal programming problem that can
be solved by a streamlined procedure.
This is a mix-integer non-linear programming problem because it involves the products of
binary variables, δriw , in Constraints (6)–(9). This type of problem can be transformed into a
mix-integer linear programming model with the aid of a series of new, non-negative variables (Fu
2007; Abraham et al. 2009), and then solved using a standard branch-and-bound technique for a
global optimum solution.

For Equation (6), the multiplication δr11 ts=r+1 (1 − δs11 ) could be replaced by the non-
negative variable u1t , and the equation could be rewritten as follows:

Nr

Qr11 = δr11 Qr11 + u1t Qt11 , ∀r ∈ N, (41)
t=r+1

t
δr11 + (1 − δs11 ) − u1t ≤ t − r, ∀ r ∈ N ; t ∈ {r + 1, . . . , Nr }, (42)
s=r+1
u1t ≤ δr11 , ∀ r ∈ N ; t ∈ {r + 1, . . . , Nr }, (43)

u1t ≤ 1 − δs11 , ∀ r ∈ N ; t ∈ {r + 1, . . . , Nr }; s ∈ {r + 1, . . . , t},
(44)
u1t ≥ 0, ∀ r ∈ N ; t ∈ {r + 1, . . . , Nr }. (45)

For Equation (7), the multiplication δr31 r−1s=t (1 − δs31 ) could be replaced by the non-negative
variable u2t , and the equation could be rewritten as follows:

r−1

Qr31 = δr31 Qr31 + u2t Qt31 , ∀r ∈ N, (46)
t=1

r−1
δr31 + (1 − δs31 ) − u2t ≤ r − t, ∀ r ∈ N ; t ∈ {1, . . . , r − 1}, (47)
s=t
u2t ≤ δr31 , ∀ r ∈ N ; t ∈ {1, . . . , r − 1}, (48)

u2t ≤ 1 − δs31 , ∀ r ∈ N ; t ∈ {1, . . . , r − 1}; s ∈ {t, . . . , r − 1}, (49)
u2t ≥ 0, ∀ r ∈ N ; t ∈ {1, . . . , r − 1}. (50)

For Equation (8), the multiplications ts=r (1 − δs11 ) and rs=t (1 − δs31 ) could be replaced by
the non-negative variables u3t and u4t , respectively, and the equation could be rewritten as follows:

Nr
r−1

Qr12 = Qr12 + u3t Qt11 + u4t Qt31 , ∀r ∈ N, (51)
t=r t=1

t
(1 − δs11 ) − u3t ≤ t − r, ∀ r ∈ N ; t ∈ {r, . . . , Nr }, (52)
s=r
u3t ≤ 1 − δs11 , ∀ r ∈ N ; t ∈ {r, . . . , Nr }; s ∈ {r, . . . , t}, (53)

u3t ≥ 0, ∀ r ∈ N ; t ∈ {r, . . . , Nr }, (54)

r
(1 − δs31 ) − u4t ≤ r − t, ∀ r ∈ N ; t ∈ {1, . . . , r − 1}, (55)
s=t
u4t ≤ 1 − δs31 , ∀ r ∈ N ; t ∈ {1, . . . , r − 1}; s ∈ {t, . . . , r}, (56)

u4t ≥ 0, ∀ r ∈ N ; t ∈ {1, . . . , r − 1}. (57)
140 J. Zhao et al.

For Equation (9), the multiplications ts=r (1 − δs11 ) and rs=t (1 − δs31 ) could be replaced by
the non-negative variables u5t and u6t , respectively, and the equation could be rewritten as follows:

Nr
r

Qr32 = Qr32 + u5t Qt11 + u6t Qt31 , ∀r ∈ N, (58)
t=r+1 t=1

t
(1 − δs11 ) − u5t ≤ t − r, ∀ r ∈ N ; t ∈ {r + 1, . . . , Nr }, (59)
s=r
u5t ≤ 1 − δs11 , ∀ r ∈ N ; t ∈ {r + 1, . . . , Nr }; s ∈ {r, . . . , t}, (60)

u5t ≥ 0, ∀ r ∈ N ; t ∈ {r + 1, . . . , Nr }, (61)

r
(1 − δs31 ) − u6t ≤ r − t, ∀ r ∈ N ; t ∈ {1, . . . , r}, (62)
s=t
u6t ≤ 1 − δs31 , ∀ r ∈ N ; t ∈ {1, . . . , r}; s ∈ {t, . . . , r}, (63)

u6t ≥ 0, ∀ r ∈ N ; t ∈ {1, . . . , r}. (64)
4. Numerical analysis
In this section, the effectiveness of the proposed model is evaluated by a series of numerical
examples. An arterial with three four-arm intersections and two internal segments is studied.
Figure 2 shows the initial layout of the arterial. The medial two lanes within the internal segments
are electable reversible lanes. The traffic demands, shown in Table 2, fluctuate during the six
different study patterns. Minimum and maximum cycle lengths are 60 and 120 s. Clearance time
for any pair of mutually incompatible traffic movements is 5.0 s. The saturation flow rate per lane
for all movements is 1900 veh h−1 ln−1 . The maximum acceptable limit of the degree of saturation
is 0.85. P1 , P2 , P3 , P4 are 10,000, 100, 1 and 0.01.
Performance of the proposed optimisation model (model 1) is compared with other three
strategies:
Model 2: Signal timing optimisation with fixed lane markings (no reversible lanes, no left-turn
restrictions, and no dynamic lane assignment).
Model 3: Reversible lanes + dynamic lane assignment strategy (no left-turn restriction).
Model 4: Reversible lanes + left-turn restriction strategy (no dynamic lane assignment) under
different traffic demand patterns.
Figure 2. Layout of the numerical example.

Table 2. Traffic demand of all patterns.
Traffic demand (pcu h−1 )

Origin Destination
arm arm Pattern 1 Pattern 2 Pattern 3 Pattern 4 Pattern 5 Pattern 6
(1,2) (1,3) 150 200 200 240 240 320
(1,2) (1,4) 300 150 150 30 30 30
(1,2) (2,2) 15 20 20 24 24 32
(1,2) (2,4) 15 20 20 24 72 32
(1,2) (3,1) 96 128 128 153 115 204
(1,2) (3,2) 12 16 16 19 14 26
(1,2) (3,4) 12 16 16 20 15 26
(1,3) (1,2) 150 100 50 60 60 80
(1,3) (1,4) 150 100 50 60 60 80
(1,3) (2,2) 120 160 160 192 192 256
(1,3) (2,4) 120 160 160 192 572 256
(1,3) (3,1) 768 1024 1024 1230 922 1640
(1,3) (3,2) 96 128 128 153 115 204
(1,3) (3,4) 96 128 128 153 115 204
(1,4) (1,2) 300 150 150 30 30 30
(1,4) (1,3) 150 200 200 240 240 320
(1,4) (2,2) 15 20 20 24 24 32
(1,4) (2,4) 15 20 20 24 72 32
(1,4) (3,1) 96 128 128 153 115 204
(1,4) (3,2) 12 16 16 20 15 26
(1,4) (3,4) 12 16 16 19 14 26
(2,2) (1,2) 15 10 5 6 6 8
(2,2) (1,3) 120 80 40 48 48 64
(2,2) (1,4) 15 10 5 6 6 8
(2,2) (2,4) 750 600 600 480 20 240
(2,2) (3,1) 120 160 160 192 576 256
(2,2) (3,2) 15 20 20 24 72 32
(2,2) (3,4) 15 20 20 24 72 32
(2,4) (1,2) 15 10 5 6 6 8
(2,4) (1,3) 120 80 40 48 48 64
(2,4) (1,4) 15 10 5 6 6 8
(2,4) (2,2) 750 900 600 1020 1020 960
(2,4) (3,1) 120 160 160 192 192 256
(2,4) (3,2) 15 20 20 24 24 32
(2,4) (3,4) 15 20 20 24 24 32
(3,1) (1,2) 96 64 32 38 38 51
(3,1) (1,3) 768 512 256 308 308 410
(3,1) (1,4) 96 64 32 38 38 51
(3,1) (2,2) 120 80 40 48 48 64
(3,1) (2,4) 120 80 40 48 48 64
(3,1) (3,2) 150 200 200 240 240 320
(3,1) (3,4) 150 200 200 240 240 320
(3,2) (1,2) 12 8 4 5 5 6
(3,2) (1,3) 96 64 32 38 38 51
(3,2) (1,4) 12 8 4 5 5 7
(3,2) (2,2) 15 10 5 6 6 8
(3,2) (2,4) 15 10 5 6 6 8
(3,2) (3,1) 150 100 50 60 60 80
(3,2) (3,4) 30 180 330 300 300 240
(3,4) (1,2) 12 8 4 5 5 7
(3,4) (1,3) 96 64 32 38 38 51
(3,4) (1,4) 12 8 4 5 5 6
(3,4) (2,2) 15 10 5 6 6 8
(3,4) (2,4) 15 10 5 6 6 8
(3,4) (3,1) 150 100 50 60 60 80
(3,4) (3,2) 30 180 330 300 300 240
142 J. Zhao et al.
Table 3. Optimised signal timing plans from the proposed model for pattern 1.
EB WB NB SB
LT TH RT LT TH RT LT TH RT LT TH RT
Intersection 1
Start of green (s) 0.0 23.9 23.9 0.0 20.4 20.4 71.5 91.9 91.9 71.5 91.9 91.9
Duration of green (s) 15.4 42.6 42.6 18.9 46.2 46.2 15.4 23.1 23.1 15.4 23.1 23.1
End of green (s) 15.4 66.5 66.5 18.9 66.5 66.5 86.9 115.0 115.0 86.9 115.0 115.0
Intersection 2
Start of green (s) 0.0 17.5 17.5 0.0 17.5 17.5 60.0 77.5 77.5 60.0 77.5 77.5
End of green (s) 12.5 55.0 55.0 12.5 55.0 55.0 72.5 115.0 115.0 72.5 115.0 115.0
Intersection 3
Start of green (s) 0.0 22.9 22.9 0.0 22.9 22.9 81.4 104.3 104.3 81.4 104.3 104.3
End of green (s) 17.9 76.4 76.4 17.9 76.4 76.4 99.3 115.0 115.0 99.3 115.0 115.0
Note: By EB, WB, NB, and SB are meant eastbound, westbound, northbound, and southbound, respectively. By LT, TH
and RT are meant left-turn, through movement and right-turn, respectively.
All the models are solved by LINGO(11.0) on an Intel® Core™ i5 2.30 GHz processor and
4.0 GB RAM, running under Windows. The computing times for all cases are less than 1 min.
The optimized signal timings and detailed lane configuration plans for all the intersections
and reversible lanes generated from the optimisation model in each pattern are illustrated in
Tables 3–8 and Figures 3–6, respectively. This information is particularly useful when designing
and operating an arterial with many intersections where each of them has multiple approach lanes
and the optimal lane channelisation plan is not straightforward to figure out through engineering
judgement.
Table 9 summarises the optimisation results from the proposed model including the maximised
common flow multipliers and the adopted optimisation strategies. Table 10 further illustrates the
performance comparison between the proposed model and the other three design methods. The
following observations can be made from Tables 9 and 10.
(1) The principal objective of the proposed model is to automatically optimise the reversible
lane strategy, the left-turn restriction strategy, the lane assignments, and the signal timings
for an arterial in a unified optimisation framework. It could change the arterial geometry
when the traffic demand varies pronouncedly and adjust the signal settings for a small
volume situation. It has, therefore, a high level of manoeuvrability. Compared with the other
Figure 3. Optimisation results of layout for patterns 1, 2 and 3.

Figure 4. Optimisation results of layout for pattern 4.
design methods, the proposed model could maintain the permitted degree of saturation for
every intersection at a minimum cost.
(2) For pattern 1, all the optimised common flow multipliers for the control area are larger than
1. This means that all the control methods can satisfy the traffic demand under the given
maximum permitted degree of saturation (0.85). In this case, which is the result optimised
by the proposed model, there is no need to use any special management strategies, and the
signal timings alone can accommodate the traffic demand.
(3) For pattern 2, 65% of the traffic moves in one direction on the internal segments. In
this case, all the control methods can still satisfy the traffic demand since the imbalance
between the traffic volumes in the two directions is small enough to be accommodated by
144 J. Zhao et al.
EB WB NB SB
Intersection 1
Start of green (s) 0.0 16.1 16.1 0.0 19.6 19.6 68.3 95.6 95.6 68.3 95.6 95.6
End of green (s) 14.6 63.3 63.3 11.1 63.3 63.3 90.6 115.0 115.0 90.6 115.0 115.0
Intersection 2
Start of green (s) 0.0 12.7 12.7 0.0 26.5 26.5 63.8 84.2 84.2 63.8 76.5 76.5
End of green (s) 21.5 58.8 58.8 7.7 58.8 58.8 71.5 115.0 115.0 79.2 115.0 115.0
Intersection 3
Start of green (s) 0.0 24.2 24.2 0.0 33.8 33.8 86.9 101.5 101.5 86.9 101.5 101.5
End of green (s) 28.8 81.9 81.9 19.2 81.9 81.9 96.5 115.0 115.0 96.5 115.0 115.0
EB WB NB SB
Intersection 1
Start of green (s) 0.0 11.0 11.0 0.0 18.8 18.8 65.2 94.1 94.1 65.2 94.1 94.1
End of green (s) 13.8 60.2 60.2 6.0 60.2 60.2 89.1 115.0 115.0 89.1 115.0 115.0
Intersection 2
Start of green (s) 0.0 9.0 9.0 0.0 34.7 34.7 62.0 83.0 83.0 62.0 73.4 73.4
End of green (s) 29.7 57.0 57.0 4.0 57.0 57.0 68.4 115.0 115.0 78.0 115.0 115.0
Intersection 3
Start of green (s) 0.0 24.2 24.2 0.0 43.5 43.5 86.9 96.7 96.7 86.9 96.7 96.7
End of green (s) 38.5 81.9 81.9 19.2 81.9 81.9 91.7 115.0 115.0 91.7 115.0 115.0
optimising signal timings alone. The result optimised by the proposed model supports the
conventional design, and no special management strategies are required.
(4) For pattern 3, there is a significant imbalance between the traffic volumes in the two direc-
tions (the ratio of the traffic demand on the major direction to that on the minor direction is
4:1). The conventional design can keep the control area under the given maximum permit-
ted degree of saturation because the total volume on the internal segment is low. Therefore,
no special management strategies are needed.
(5) For pattern 4, the traffic demand increases about 20% over pattern 3. The optimisation result
of model 2 shows that the traffic demand imbalance exceeds the signal timing adjustment
effect. Therefore, to balance the degree of saturation of the two travel directions, the
‘reversible lanes + dynamic lane’ assignment strategy is selected as the optimisation
result of the proposed model. The reversible lanes operation is justified for the situation
EB WB NB SB
Intersection 1
Start of green (s) 0.0 11.5 11.5 0.0 15.9 15.9 69.7 100.5 100.5 69.7 100.5 100.5
End of green (s) 10.9 64.7 64.7 6.5 64.7 64.7 95.5 115.0 115.0 95.5 115.0 115.0
Intersection 2
Start of green (s) 0.0 10.0 10.0 0.0 28.5 28.5 60.0 85.0 85.0 60.0 70.0 70.0
End of green (s) 23.5 55.0 55.0 5.0 55.0 55.0 65.0 115.0 115.0 80.0 115.0 115.0
Intersection 3
Start of green (s) 0.0 28.5 28.5 0.0 43.2 43.2 86.5 97.4 97.4 86.5 97.4 97.4
End of green (s) 38.2 81.5 81.5 23.5 81.5 81.5 92.4 115.0 115.0 92.4 115.0 115.0
EB WB NB SB
Intersection 1
Start of green (s) 0.0 11.5 11.5 0.0 15.8 15.8 69.6 100.5 100.5 69.6 100.5 100.5
End of green (s) 10.8 64.6 64.6 6.5 64.6 64.6 95.5 115.0 115.0 95.5 115.0 115.0
Intersection 2
Start of green (s) 0.0 9.9 9.9 0.0 33.9 33.9 60.7 85.1 85.1 60.7 70.7 70.7
End of green (s) 28.9 55.7 55.7 4.9 55.7 55.7 65.7 115.0 115.0 80.1 115.0 115.0
Intersection 3
Start of green (s) 0.0 28.5 28.5 0.0 43.2 43.2 86.5 97.4 97.4 86.5 97.4 97.4
End of green (s) 38.2 81.5 81.5 23.5 81.5 81.5 92.4 115.0 115.0 92.4 115.0 115.0
where the traffic flow is relatively high and the difference of the volume between the major
and minor traffic directions is significant. Since the number of approach and exit lanes is
changed, the lane markings at the intersection have been adjusted accordingly.
(6) For pattern 5, using the traffic demand in pattern 4 as a benchmark, the left-turn proportion
in arm (2,3) is raised from 10% to 30%. The proposed model can accommodate this kind
of traffic demand variation well using the dynamic lane assignment strategy. It can be seen
that the combined management strategy of reversible lanes and dynamic lane assignment
has a high level of manoeuverability. This can accommodate the traffic demand variations
of both the imbalance of the two directions and the turning proportion.
(7) For pattern 6, the total volume is too high to be accommodated by control strategies other
than the ‘reversible lanes + left-turn restriction’ strategy. Therefore, it should be selected
as the optimisation result that is the same as the proposed model. It can be seen, however,
146 J. Zhao et al.
EB WB NB SB
Intersection 1
Start of green (s) 0.0 17.2 17.2 0.0 16.3 16.3 72.4 101.7 101.7 72.4 101.7 101.7
End of green (s) 11.3 67.4 67.4 12.2 67.4 67.4 96.7 115.0 115.0 96.7 115.0 115.0
Intersection 2
Start of green (s) – 0.0 0.0 – 0.0 0.0 59.8 91.6 91.6 59.8 71.5 71.5
Duration of green (s) – 54.8 54.8 – 54.8 54.8 6.7 23.4 23.4 26.8 43.5 43.5
End of green (s) – 54.8 54.8 – 54.8 54.8 66.5 115.0 115.0 86.6 115.0 115.0
Intersection 3
Start of green (s) 0.0 30.0 30.0 0.0 59.2 59.2 91.3 102.5 102.5 91.3 102.5 102.5
End of green (s) 54.2 86.3 86.3 25.0 86.3 86.3 97.5 115.0 115.0 97.5 115.0 115.0
Table 9. Optimised common flow multiplier for the example.
Pattern 1 2 3 4 5 6
Total volume on each internal segment (veh h−1 ) 3000 3000 2500 3000 3000 4000
Ratio of a major to minor traffic count on the 1:1 2:1 4:1 4:1 4:1 4:1
internal segment
Left-turn proportion in arm (1,1) 0.1 0.1 0.1 0.1 0.1 0.1
Adoption of the signal timing optimisation Yes Yes Yes Yes Yes Yes
Adoption of the reversible lanes No No No Yes Yes Yes
Adoption of the dynamic lane assignment No No No Yes Yes Yes
Adoption of the left-turn restriction No No No No No Yes
Maximum degree of saturation 0.85 0.85 0.85 0.85 0.85 0.85
Maximum μ 1.12 1.04 1.08 1.12 1.09 1.02
that the application of this management strategy has many limitations, including forcing
the left-turn traffic to make a detour. This, in turn, may cause extra driving and increase
the total vehicle-miles travelled in the network as well as shift the critical intersection. The
nearby intersections, therefore, must have low saturation to accommodate the additional
traffic flow.
In addition to the deterministic capacity analysis mentioned above, the simulation package,
VISSIM, was used as an unbiased evaluator to capture the stochastic variation of the traffic flow
and evaluate the performance of the optimisation results. Two performance indices, the total
throughput and the average vehicular delay, were selected. The analysis and comparison results
of the two performance indices are shown in Figures 7 and 8, respectively.
It can be seen that in patterns 1, 2, and 3, the control plans were all comparable in terms of
total throughput. Though model 4 outperformed the other designs in average delay, the advantage
was insignificant since signal timing optimisation could have also ensured that the intersection
would perform reasonably well. In patterns 4 and 5, the signal timing optimisation (model 2)
Table 10. Comparison between different control strategies.
Pattern 1 2 3 4 5 6
Model 1: proposed Degree of saturation of intersection 1 0.691 0.588 0.570 0.659 0.638 0.850
model Degree of saturation of intersection 2 0.850 0.850 0.850 0.850 0.850 0.773
Degree of saturation of intersection 3 0.595 0.680 0.707 0.723 0.700 0.827
Number of prohibited movements 0 0 0 0 0 2
Number of direction reversed lanes 0 0 0 2 2 2
Maximum μ 1.122 1.035 1.077 1.122 1.087 1.023
Model 2: signal Degree of saturation of intersection 1 0.691 0.588 0.570 0.573 0.524 0.640
timing Degree of saturation of intersection 2 0.850 0.850 0.850 0.850 0.850 0.850
optimisation Degree of saturation of intersection 3 0.595 0.680 0.707 0.739 0.677 0.788
alone Number of prohibited movements 0 0 0 0 0 0
Maximum μ 1.122 1.035 1.077 0.975 0.894 0.821
Model 3: reversible Degree of saturation of intersection 1 0.691 0.588 0.570 0.659 0.638 0.750
lanes + dynamic Degree of saturation of intersection 2 0.850 0.850 0.850 0.850 0.850 0.850
lane assignment Degree of saturation of intersection 3 0.595 0.680 0.707 0.723 0.700 0.777
Maximum μ 1.122 1.035 1.077 1.122 1.087 0.961
Model 4: reversible Degree of saturation of intersection 1 0.850 0.793 0.712 0.809 0.585 0.850
lanes + left-turn Degree of saturation of intersection 2 0.810 0.850 0.850 0.850 0.621 0.773
restriction Degree of saturation of intersection 3 0.831 0.825 0.834 0.834 0.850 0.827
Maximum μ 1.329 1.256 1.270 1.294 0.937 1.023
Figure 7. Throughput analysis.
could not accommodate the traffic demand. In this case, the proposed model selected model
3, the optimisation design scheme, which clearly outperformed model 2 with significantly lower
average delays (20.8% and 41.7% reduction for patterns 4 and 5, respectively). In pattern 6, neither
model 2 nor 3 designs could accommodate the traffic demand. The proposed model, therefore,
selected model 4, the optimisation design scheme, which clearly outperformed models 2 and 3
with significantly lower average delays (46.8% and 23.6% reduction, respectively).
The simulation analysis also indicated that the effectiveness of special traffic management
strategies such as lane reversion and turning restriction is very limited when the intersection can
148 J. Zhao et al.
Figure 8. Delay analysis.
perform reasonably well. Therefore, it is not recommended to consider special traffic management
strategies when the conventional design can accommodate the traffic demand.
5. Conclusion
This paper presented a lane-based optimisation model for the integrated design of an arterial. It
was formulated using a multi-objective mix-integer non-linear programming problem to simul-
taneously optimise the reversible lanes, left-turn restriction, lane assignments, and signal timing.
The problem was transformed into a single-objective integer linear programming model using
a streamlined procedure and introducing a series of new non-negative variables. This was then
solved by a standard branch-and-bound technique.
Extensive numerical analysis was conducted to evaluate the performance of the proposed design
under different traffic demand patterns, from which the following conclusions could be drawn.
(1) The optimisation design scheme obtained by the proposed model could automatically
select the best control strategies according to the traffic demand and the layout situation.
(2) In low-demand or little imbalanced flow distribution between two directions patterns, there
is no need to use any special management strategies, and signal timing could accommodate
these small volume fluctuations quite well.
(3) The combined management strategy of reversible lanes and dynamic lane assignment has
a high level of manoeuvrability that can accommodate the traffic demand variations of
both the imbalance of the two directions and the turning proportion.
(4) The application of the combined management strategy of reversible lanes and left-turn
restriction has many limitations. This option must be carefully considered to prevent new
traffic congestion that may be caused by the additional detour traffic.
Although this approach provides an effective method to design reversible lanes, reserve
capacity by itself may not form a complete set of performance measurement. It can be
extended to include considerations of delay as an objective. Moreover, deterministic traffic
demand is assumed while incorporating stochastic demand is also a valuable extension.
It is assumed that the segment of the arterial, in this paper, is sufficiently long. In a real
application, overflowing queue problems should be considered. Considering the variable nature
of reversible lanes, it is not recommended for practical implementation unless the conventional
design cannot accommodate the traffic demand. Moreover, the effective traffic control is vital to
the safe operation of reversible lanes. In addition to guiding vehicles into and out of the vicinity
of the reversible segments, control features must be used to effectively communicate critical
information, such as where and when reversible operations begin and end, and what lanes are
available to drivers (Wolshon and Lambert 2004). And suitable clearance time should be set
during the transition of reversible lanes.
Funding
The research is supported by the National Natural Science Foundation of China [Grant Nos. 51238008 and
51178345], and the Fundamental Research Funds for the Central Universities.
References
AASHTO (American Association of State Highway and Transportation Officials). 2011. A Policy on Geomet-
ric Design of Highways and Streets. 6th ed. Washington, DC: American Association of State Highway
and Transportation Officials.
Abraham, A., A.-E. Hassanien, P. Siarry, and A. Engelbrecht. 2009. Foundations of Computational
Intelligence. Vol. 3. Berlin: Springer.
Agent, K. R., and J. D. Clark. 1980. Evaluation of Reversible Lanes (Nicholasville Road; Lexington, KY).
Lexington: Kentucky Department of Transportation.
Dorsey, R. T. 1948. “The Use of the Off-Centerlane Movement in Los Angeles.” Traffic Quarterly 2 (3):
291–302.
FGSV (Road and Transportation Research Association Working Group Traffic Control and Traffic Safety).
2006. Guidelines for Traffic Signals (RiLSA). [In Chinese.] Translated by K.P. Li. Beijing: China
Architecture & Building Press.
Ford, G., R. Henk, and P. Barricklow. 2000. Interstate Highway 37 Reverse-Flow Analysis. College Station:
Texas Transportation Institute.
Fu, J. L. 2007. Operational Research Methodology and Model. [In Chinese.] Shanghai: Fudan University
Press.
Gallivan, S., and B. G. Heydecker. 1988. “Optimising the Control Performance of Traffic Signals at a Single
Junction.” Transportation Research Part B: Methodological 22 (5): 357–370.
Gillis, R. D. 1990. “Unlocking Arena Gridlock.” Civil Engineering 60 (2): 43–45.
ITE (Institute of Transportation Engineers). 2009. Traffic Engineering Handbook. 6th ed. Washington, DC:
Institute of Transportation Engineers.
JSCE (Japan Society of Civil Engineers). 1989. Handbook of Civil Engineering. [In Japanese.] 4th ed. Tokyo:
GihodoShuppan.
Karoonsoontawong, A., and D. Lin. 2011. “Time-Varying Lane-Based Capacity Reversibility for Traffic
Management.” Computer-Aided Civil and Infrastructure Engineering 26 (8): 632–646.
Lam, W. H. K., A. C. K. Poon, and G. K. S. Mung. 1997. “Integrated Model for Lane-Use and Signal-Phase
Designs.” Journal of Transportation Engineering 123 (2): 114–122.
Lambert, L., and B. Wolshon. 2002. Analysis of Reverse Flow Traffic Operations Phase I: Urban Sporting
Event Measurement and Evaluation. Baton Rouge: Louisiana State University.
PBS&J. 2000. Reverse Lane Standards and Its Strategies: Southeast United States Hurricane Study Technical
Memorandum. Tallahassee, FL: PBS&J.
TRB (Transportation Research Board). 2000. Highway Capacity Manual 2000. Washington, DC: Trans-
portation Research Board.
Tuydes, H. 2005. “Network Traffic Management under Disaster Conditions.” PhD diss., Northwestern
University.
Tuydes, H., and A. Ziliaskopoulos. 2004. “Network Re-design to Optimize Evacuation Contraflow.” Paper
presented at the 83rd annual meeting of the Transportation Research Board, Washington, DC, January
11–15.
Upchurch, J. 1975. “Reversible Flow on a Six Lane Urban Arterial.” Traffic Engineering 45 (12): 11–14.
Urbina, E., and B. Wolshon. 2003. “National Review of Hurricane Evacuation Plans and Policies: A Com-
parison and Contrast of State Practices.” Transportation Research Part A: Policy and Practice 37 (3):
257–275.
Wolshon, B. 2001. “One-Way-Out: Contraflow Freeway Operation for Hurricane Evacuation.” Natural
Hazards Review 2 (3): 105–112.
Wolshon, B. 2002. “Planning for the Evacuation of New Orleans.” ITE Journal 72 (2): 44–49.
150 J. Zhao et al.
Wolshon, B., and L. Lambert. 2004. Convertible Roadways and Lanes: A Synthesis of Hignwaypractice.
Washington, DC: Transportation Research Board.
Wolshon, B., and L. Lambert. 2006. “Planning and Operational Practices for Reversible Roadways.” ITE
Journal 76 (8): 38–43.
Wong, C. K., and B. G. Heydecker. 2011. “Optimal Allocation of Turns to Lanes at an Isolated Signal-
Controlled Junction.” Transportation Research Part B: Methodological 45 (4): 667–681.
Wong, C. K., and S. C. Wong. 2002. “Lane-Based Optimization of Traffic Equilibrium Settings for Area
Traffic Control.” Journal of Advanced Transportation 36 (3): 349–386.
Wong, C. K., and S. C. Wong. 2003a. “Lane-Based Optimization of Signal Timings for Isolated Junctions.”
Transportation Research Part B: Methodological 37 (1): 63–84.
Wong, C. K., and S. C. Wong. 2003b. “A Lane-Based Optimization Method for Minimizing Delay at Isolated
Signal-Controlled Junctions.” Journal of Mathematical Modelling and Algorithms 2 (4): 379–406.
Wu J., H. Sun, Z. Gao, and H. Zhang. 2009. “Reversible Lane-Based Traffic Network Optimization with an
Advanced Traveller Information System.” Engineering Optimization 41 (1): 87–97.
Xie, C., D. Lin, and S. Waller. 2010. “A Dynamic Evacuation Network Optimization Problem with
Lane Reversal and Crossing Elimination Strategies.” Transportation Research Part E: Logistics and
Transportation Review 46 (3): 295–316.
Xie, C., and M. A. Turnquist. 2011. “Lane-Based Evacuation Network Optimization: An Integrated
Lagrangian Relaxation and Tabu Search Approach.” Transportation Research Part C: Emerging
Technologies 19 (1): 40–63.
Yang, H., and M. G. H. Bell. 1998. “Models and Algorithms for Road Network Design: A Review and Some
New Developments.” Transport Reviews 18 (3): 257–278.
Yu, X., and P. D. Prevedouros. 2013. “Left-Turn Prohibition and Partial Grade Separation for Signalized
Intersections: Planning-Level Assessment.” Journal of Transportation Engineering 139 (4): 399–406.
Zhang, H., and Z. Gao. 2007. “Two-Way Road Network Design Problem with Variable Lanes.” Journal of
Science and Systems Engineering 16 (1): 50–61.
Zhao, J., W. Ma, H. Michael Zhang, and X.Yang. 2013a. “Increasing the Capacity of Signalized Intersections
by Dynamically Using Exit-Lanes for Left-Turn Traffic.” Paper presented at the 92nd annual meeting
of the Transportation Research Board, Washington, DC, January 13–17.
Zhao, J., W. Ma, H. Michael Zhang, and X. Yang. 2013b. “Two-Step Optimization Model for Dynamic Lane
Assignment at Isolated Signalized Intersections.” Paper presented at the 92nd annual meeting of the
Transportation Research Board, Washington, DC, January 13–17.

zhao2014

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

zhao2014

Uploaded by

Copyright:

Available Formats

Transportmetrica B: Transport Dynamics, 2014

Vol. 2, No. 2, 130–150, http://dx.doi.org/10.1080/21680566.2014.908751

Integrated design and operation of urban arterials with reversible lanes

∗ Corresponding author. Email: mawanjing@tongji.edu.cn

© 2014 Hong Kong Society for Transportation Studies Limited

2. Model configuration and assumptions

Figure 1. Configuration of the integrated model.

(1) Traffic demand distribution in the target arterial is assumed to be given.

3. Integrated optimisation model

3.2. Objective function

where P1 , P2 , P3 , P4 are priority coefficients for the four control objectives, P1 P2 P3 P4 ;

Mxriwk ≥ qriwk ≥ 0, ∀ r ∈ N ; i ∈ Ar ; w ∈ Ti ; k ∈ {1, . . . , nriaf + nriac }, (4)

3.3.2. Lane assignment constraints

nr(r+1),(r+1)r = nr(r+1) + n(r+1)r , ∀r ∈ N, (13)

1 − xriw(k+1) ≥ xriw k , ∀ r ∈ N ; i ∈ Ar ; w ∈ {1, 2}; w

(6) Number of lanes assigned to prohibited movements should be equal to 0

δriw ≥ xriwk , ∀ r ∈ N ; i ∈ Ar ; w ∈ Ti ; k ∈ {1, . . . , nriaf + nriac }. (22)

3.3.3. Signal timing constraints

M(1 − xriwk ) ≥ Grik − griw ≥ −M(1 − xriwk ),

3.3.4. Flow ratio constraints

3.3.5. Degree of saturation constraints

3.4. Solution algorithm

u1t ≤ δr11 , ∀ r ∈ N ; t ∈ {r + 1, . . . , Nr }, (43)

u2t ≤ δr31 , ∀ r ∈ N ; t ∈ {1, . . . , r − 1}, (48)

u3t ≤ 1 − δs11 , ∀ r ∈ N ; t ∈ {r, . . . , Nr }; s ∈ {r, . . . , t}, (53)

u4t ≤ 1 − δs31 , ∀ r ∈ N ; t ∈ {1, . . . , r − 1}; s ∈ {t, . . . , r}, (56)

u5t ≤ 1 − δs11 , ∀ r ∈ N ; t ∈ {r + 1, . . . , Nr }; s ∈ {r, . . . , t}, (60)

u6t ≤ 1 − δs31 , ∀ r ∈ N ; t ∈ {1, . . . , r}; s ∈ {t, . . . , r}, (63)

Figure 2. Layout of the numerical example.

Table 2. Traffic demand of all patterns.

Traffic demand (pcu h−1 )

Figure 3. Optimisation results of layout for patterns 1, 2 and 3.

Figure 4. Optimisation results of layout for pattern 4.

Figure 5. Optimisation results of layout for pattern 5.

Figure 6. Optimisation results of layout for pattern 6.

Table 9. Optimised common flow multiplier for the example.

Table 10. Comparison between different control strategies.

Figure 7. Throughput analysis.

Figure 8. Delay analysis.

You might also like