energies

Article
Optimizing Multi Cross-Docking Systems with a Multi-Objective
Green Location Routing Problem Considering Carbon Emission
and Energy Consumption
Ieva Meidute-Kavaliauskiene 1, * , Nihal Sütütemiz 2 , Figen Yıldırım 3 , Shahryar Ghorbani 4
and Renata Činčikaitė 1

1 Department of Business Technologies and Entrepreneurship, Vilnius Gediminas Technical University,

Sauletekio al. 11, 10223 Vilnius, Lithuania; [email protected]
2 Department of Management Information Systems, University of Sakarya, Sakarya 54050, Turkey;
[email protected]
3 Department of International Trade, Istanbul Commerce University, Istambul 34445, Turkey;
[email protected]
4 Department of Production Management, University of Sakarya, Sakarya 54050, Turkey;
[email protected]
* Correspondence: [email protected]

Abstract: Cross-docking is an excellent way to reduce the space required to store goods, inventory
management costs, and customer order delivery time. This paper focuses on cost optimization,
scheduling incoming and outgoing trucks, and green supply chains with multiple cross-docking. The
three objectives are minimizing total operating costs, truck transportation sequences, and carbon





 emissions within the supply chain. Since the linear programming model is an integer of zero and one
Citation: Meidute-Kavaliauskiene, I.;
and belongs to NP-hard problems, its solution time increases sharply with increasing dimensions.
Sütütemiz, N.; Yıldırım, F.; Ghorbani, Therefore, the non-dominated sorting genetic algorithm-II (NSGA-II) and the multi-objective particle
S.; Činčikaitė, R. Optimizing Multi swarm optimization (MOPSO) were used to find near-optimal solutions to the problem. Then, these
Cross-Docking Systems with a algorithms were compared with criteria such as execution time and distance from the ideal point,
Multi-Objective Green Location and the superior algorithm in each criterion was identified.
Routing Problem Considering
Carbon Emission and Energy Keywords: non-dominated sorting genetic algorithm-II (NSGA-II); multi-objective particle swarm
Consumption. Energies 2022, 15, 1530. optimization (MOPSO); cross-docking
https://doi.org/10.3390/en15041530

Academic Editors: Bernard Zi˛ebicki

and Edyta Bielińska-Dusza
1. Introduction
Received: 21 December 2021
Accepted: 16 February 2022
Logistics strategy is an essential advantage for supply chain management operations
Published: 18 February 2022
that requires centralized planning and adjustment of procedures to reduce costs and
increase customer satisfaction [1,2]. Classic supply chain management for transporting
Publisher’s Note: MDPI stays neutral
goods from suppliers to the distribution centre requires warehousing for storage [3] and
with regard to jurisdictional claims in
then packaging and shipping to customers [4], according to their demand [5]. This form
published maps and institutional affil-
of supply chain management has high inventory maintenance costs and human resources
iations.
costs [6,7]. In other words, there are four stages of receiving, storage, order selection,
and sending in traditional warehouses [8]. Still, in the cross-dock, the storage stage,
one of the main obstacles of inventory systems, is removed [9,10]. Therefore, it reduces
Copyright: © 2022 by the authors.
inventory and increases customer satisfaction [11]. Cross-dock is a logistic method that
Licensee MDPI, Basel, Switzerland. distributes products directly from a supplier or factory [12]. The cross-dock method ensures
This article is an open access article timely transportation and better coordination between supply and demand [13]. In recent
distributed under the terms and years, one of the issues facing the supply chain was clean production and environmental
conditions of the Creative Commons protection [5]. One of the critical issues of clean output is using supply chains that pay
Attribution (CC BY) license (https:// attention to environmental issues in all supply chain processes [14,15].
creativecommons.org/licenses/by/ In this research, we investigate the operational cross-docking decisions that concern
4.0/). the optimization of short-term decisions such as CO2 emission and energy consumption

Energies 2022, 15, 1530. https://doi.org/10.3390/en15041530 https://www.mdpi.com/journal/energies

Energies 2022, 15, 1530 2 of 24

that are directly related to the transit of products from inbound trucks to outbound trucks.
We focus on the optimization of the truck to door assignment problem, which is one of the
key issues in cross-docking. It attempts to find the optimal assignment for each incoming
inbound or outbound truck to the appropriate inbound or outbound dock according to the
cross-dock characteristics. When the problem of truck-to-door assigning is mentioned, two
different questions should be answered: at which door and in which order the trucks should
be docked. The first question was answered by many researchers, but the second question
receives less attention and this research tries to investigate the answer to this question.
The cross-docking system targets to reduce the cost of inventory holding and minimize
the delivery time from suppliers to retailers in the supply chain. However, this paper
addresses the cross-dock selection and optimization problem, which was first considered
with the vehicle routing problem (VRP) by Maknoon and Laporte (2017), to make optimal
decisions on route construction and freight consolidation cross-docks. In other words,
selecting cross-docks as freight consolidation points leads to gaining economies of scale.
The products are processed and transported to the customers by at least one cross-dock in
the CDS problem. It has a vast application in production and retail companies and logistic
service providers that handle various shipments on large networks.
Dondo and Cerda (2014) took a different approach to applying the time window so
that a variable was used to determine the time window interval [16]. If there is a time
window, this variable will be greater than zero for error; otherwise, it will be zero [13].
Mohtashami et al. (2015), in a study to provide and optimize a multi-objective mathematical
model in the cross-dock and the entire supply chain, assumed that the supplier can also
send the product directly to the customer and send it to the cross-dock [15]. On the other
hand, multiple objectives, such as minimizing the total transport time of trucks, total cost,
number of shipments, and model solving with NSGA-II and MOPSO with unsuccessful
sorting and comparing the results of the above two algorithms, were included in their
research. Mohtashami (2015) presented a new dynamic genetic algorithm based on the
vehicle scheduling method in the cross-dock system to reduce the operation time. This
algorithm assumed the existence of a temporary warehouse for receiving and sending
goods, frequent entry and exit of trucks, and two types of chromosomes determined for
incoming and outgoing trucks, providing a shorter solution for the operating time [17].
Ponboon et al., 2016, evaluated the cost structure and the impact of parameters on location
issues and time windows in the distribution network. Nine scenarios of the Osaka distri-
bution network of freight cargo with different warehouse locations were tested with two
criteria, warehouse size and vehicle size. Therefore, using a large warehouse with a large
vehicle will reduce costs. Warehouse features, vehicles, and shipping information were also
discussed in detail [18]. Gomes et al. (2018) conducted bibliometric research in warehouse
management systems from 2006 to 2016 [19].
Birim (2016) investigated the vehicle routing problem with cross-dock, considering
heterogeneous vehicles with variable capacity and seeking to minimize the objective func-
tion of the total cost of transportation. The problem was solved with a simulated annealing
algorithm, and the best answer is compared [20]. Yin and Chuang (2016) emphasized that
we must focus on the green supply chain for sustainable development in the supply chain
and pay attention to the friendly environment and business values. This paper deals with
the minimum cost of green vehicle routing in transporting final products from suppliers to
customers through a cross-dock with limited carbon dioxide emissions. Load management
for long distances with the lowest cost and carbon dioxide emissions determine high fuel
efficiency with the tabu search algorithm [21]. Wisittipanich and Hengmeechai (2017) said
that one of the fundamental operational management problems is scheduling trucks in as-
signing them to cross-dock doors and arranging all domestic and foreign trucks for loading
and unloading. This paper presents a mathematical model of integer hybrid programming
for allocating and setting trucks in a multi-door cross-docking system. The purpose of this
model is to minimize the total operation time. Then, a modified particle swarm algorithm
is proposed and optimized with special unique designs, coding, and decoding to solve the
Energies 2022, 15, 1530 3 of 24

problem of truck scheduling in a cross-docking system [16]. Zulaga et al. (2017) showed
that this type of dock has results such as reducing cost time and improving management
information in reverse processes. Sensitivity analysis of this model helps companies en-
hance their competitive position by providing flexibility concerning products, reducing
the likelihood of product returns from the secondary market, and combining product
returns and cross-dock costs relative to traditional warehousing [14]. Mohtashami and
Najafabadi (2014) examined the reduction of operating time in the entire supply chain
in their article. In the proposed model, the incoming trucks move directly towards the
customers or the cross-dock in the first stage after loading the products from the suppliers.
The products are unloaded in the cross-dock and loaded in the output trucks in the second
stage. Then, the products are transferred to the customers [22].
Table 1 summarize the literature regarding optimizing multi cross-docking systems
(MCDS) and compare the existing problems with the problem proposed in this paper.
Table 1. Summary of literature regarding MCDS and its different variants.

Crossdocks
Capacity in
Multiple Cross-Docks

Solution
Method
Type of
Vehicle

Multiple Objectives
Time Windows
References

Homogeneous Heterogeneous Limited Unlimited Exact Heuristic Metaheuristic

[23] * * * *
[24] * * * * *
[25] * * * *
[26] * * * *
[27] * * * * *
[28] * * * * *
[29] * * *
[30] * * * *
[31] * * *
[32] * * * * *
[9] * * * * * *
[33] * * * * * *
[34] * * * *
[11] * * * * * *
[35] * * * * *
Current
* * * * * * *
research

Since cross-docks and related transportations have significant roles in increasing the ef-
ficiency of large-scale supply chain distribution networks, considering the other real-world
essential factors such as reliability and pollution is attractive in designing or redesigning
this network. Reliability maximization leads to customer satisfaction and earning a lot of
market shares, whereas pollution minimization creates eco-friendly industries.
The research conducted was generally related to cross-dock and focused on optimizing
the timing of trucks and the cost of transportation inside the dock. In this research, by
considering several cross-docks and concentrating on green management with a norm
Energies 2022, 15, 1530 4 of 24

time window approach for customers, issues related to out-of-cross-dock and the entire
supply chain operations are also addressed. This study focuses on optimizing the cost of
transporting trucks by locating the cross-dock, the sequence of transporting trucks, and the
amount of carbon dioxide emissions in the environment simultaneously throughout the
supply chain. In summary, the contributions of this research are described as follows.
• Developing a novel mathematical model to integrate CDS results in a comprehensive
problem with a great application in industries;
• Making the problem closer to the real-world conditions by considering the effects
of GHG emission in the transportation process and calculating energy consumption
dependent on traffic time;
• Investigating the total cost minimization related to transportation, minimizing truck
transportation sequences and carbon emissions through cross-docking simultaneously;
• Designing high-quality algorithms, including NSGA-II and MOPSO, efficiently to
solve the large-sized problem.

2. Problem Statements
In this paper, three objective functions, including minimizing the total cost of operation,
truck transport sequence, and carbon dioxide emissions, are examined. The proposed model
also examines the relationship between (cross-dock–suppliers), (cross-dock–customers),
(suppliers–customers), (suppliers–suppliers), (customers–customers), and (cross-dock–
cross-dock) in terms of fuzzy scheduling with a norm time window approach for truck
transport and reducing greenhouse gas emissions. The assumptions of the model are:
• All incoming and outgoing trucks are available in zero time;
• Homogeneous vehicles with different capacities;
• Considering the period of customer service of norm type (having the earliest service
start time and the latest service start time);
• Existence of several cross-docks so that suppliers choose one of the docks to send the
goods considering the minimum cost;
• Ability to connect suppliers, docks, and customers with each other;
• The type and number of products supplied by suppliers, as well as the type and
number of customer demand, are clear and constant;
• In the transport sequence, a truck can load products from more than one supplier and
unload products from more than one customer;
• One or more suppliers may meet a customer’s requirements;
• The type and quantity of products transported by incoming trucks must be equal to
the demands of customers.
To better understand and explain the research problem, the flow of operations in the
supply chain is drawn in Figure 1. This figure considers several cross-dock distribution
and transportation operation flows and several customers and suppliers. Vehicles with
different capacities are also used to transport products in the supply chain.

Figure 1. Operation process in the cross-dock.

Energies 2022, 15, 1530 5 of 24

After loading the products from the suppliers, the incoming trucks move directly
to the customers, move to another supplier, move to one of the cross-docks, and unload
the products in the cross-dock. They are then loaded onto output trucks and delivered to
customers [36]. It should also be noted that a truck can load products from more than one
supplier, and a truck can move to more than one customer and unload products between
them. Three trucks with different capacities are used to transport the products.

2.1. Mathematical Model

The main variables and important parameters in the model and algorithms are pre-
sented in Table 2:

Table 2. Sets, parameters, and decision variables of the model.

Sets
S Set of suppliers (1, . . . ,S)
C Set of customers (1, . . . ,C)
kin Set of incoming trucks (Receiving) (1, . . . ,K)
kout Set of outgoing trucks (Sending) (1, . . . ,K’)
G Set of products type (order) (1, . . . ,G)
Cd Set of Cross-docks (1, . . . ,Z)
Parameters
T Transportation time
Time of entry or exit of incoming truck (i) from supplier (s), while loading the order of type (G) and moving towards
AiscG
the customer (c)
j
AcdcG When the incoming truck (j) enters to the customer (c) from the cross-dock (cd) while loading the order of type (G)
TiscG Transportation time of incoming truck (i) from supplier (s) to customer (c) while loading the order of type (G)
j
TcdcG Transport time of outgoing truck (j) from cross-dock (cd) to customer (c), while loading the order of type (G)
ULiscG Offloading time of each product type (G) from input truck (i) to customer (c)
ULiscG Offloading time of each product type (G) from output truck (j) to customer (c)
DcG Customer order (c) of type (G) goods
PGs Production rate of product with type (G)
wG Weight of product with type (G)
di Demand of customer (i)
[ei , Ei , Li , li ] Time window interval of customer (i)
γi =
Early arrival time of incoming or outgoing truck in the time window
Ei − AiscG
θi =
Incoming or outgoing truck late arrival time
AiscG − Li
PEkout Penalty for delay or early arrival of a vehicle exiting the cross-dock (i) for the customer (c)
PEkin Penalty for delay or early arrival of the incoming vehicle from the supplier (i) to the customer (i)
wi Vehicle waiting time at the location of customer (c)
oi Cost of reopening the cross-dock (i)
i Capacity of incoming truck (i)
Q
j Capacity of output truck (i)
Q
dscd Distance between supplier (s) and cross-dock (cd)
dsc Distance between supplier (s) and customer (c)
Energies 2022, 15, 1530 6 of 24

Table 2. Cont.

Sets
dcdc Distance between cross-dock (cd) and customer (c)
Ckin The fuel conversion rate of the unloaded incoming truck to carbon dioxide
The difference between the conversion rate of the fuel of an incoming truck with a load of one unit of product or
C/kin
more with the same truck without a load of carbon dioxide
The difference between the conversion rate of the fuel of an outgoing truck with a load of one unit of product or
C/kout
more with the same truck without a load of carbon dioxide
Ckout Conversion rate of unloaded truck fuel into CO2
gisc Incoming truck fuel consumption rate for travel between supplier (s) and customer (c) without load
giscd The fuel consumption rate of incoming trucks for travel between supplier (s) and cross-dock (cd) without load
Incoming truck fuel consumption rate for travel between the source supplier (s) and the destination supplier
gism sn
without load
Incoming truck fuel consumption rate for travel between the cross-dock (cd) of origin and the cross-dock of
gicdm cdn
destination without cargo
j
gcm cn Outgoing truck fuel consumption rate for travel between origin and destination customer (c) without load
j
gcdc Outgoing truck fuel consumption rate for travel between cross-dock (cd) and customer (c) without load
The difference in the fuel consumption rate of an incoming truck traveling with one product or more with the same
Qkin
unladen truck
The difference in the fuel consumption rate of an outgoing truck traveling with one product or more with the same
Qkout
unladen truck
fscdG The number of product types (G) the truck carries between supplier (s) and cross-dock (cd)
fscG The number of product types (G) the truck carries between supplier (s) and the customer (c)
fcdcG The number of product types (G) the truck carries between cross-dock (cd) and customer (c)
fsm sn G The number of product types (G) the truck carries between source supplier and destination supplier
fcm cn G The number of product types (G) the truck carries between source customer and destination customer
fcdm cdn G The number of product types (G) the truck carries between source cross-dock and destination cross-dock
E Big number
Variables
qGs The number of product types (G) that are loaded from the supplier (s) into the input truck (i)
qGcd The number of product types (G) that are loaded from the cross-dock (cd) into the output truck (j)
qGc The number of product types (G) that are unloaded from the incoming truck (i) at the customer’s location
qGsn The number of product types (G) that are unloaded from the input truck (i) at the destination supplier sn
qGcdn The number of product types (G) that are unloaded from the output truck (j) at the destination cdn

I f incoming truck i moves f rom cross − dock cd
1

XicdsG to supplier S while loading the order o f type G
 0
otherwise

I f incoming truck i moves f rom supplier Sm
1

Xism sn G to destination Supplier Sn while loading the order o f type G
 0
otherwise

I f incoming truck i moves f rom supplier S
1

XiscdG to cross − dock cd while loading order o f type G
 0
otherwise

I f incoming truck i moves f rom supplier S
1

XiscG to customer C while loading the order o f type G
 0
otherwise
Energies 2022, 15, 1530 7 of 24

Table 2. Cont.

Sets

I f incoming truck i moves f rom customer Cm
1

Xicm cn G to destination customer Cn while loading the order o f type G
 0
otherwise

I f outgoing truck j moves f rom cross − dock cd
1

j
XcdcG to customer C while loading the order o f type G
 0
otherwise

I f incoming truck i moves f rom customer C
1

XiccdG to cross − dock cd while o f f loading type G order
 0
otherwise

I f incoming truck j moves f rom customer C
1

j
XccdG to cross − dock cd while o f f loading type G order
 0
otherwise

I f incoming truck i moves f rom source cross − dock cdm
1

Xicdm cdn G to destination cross − dock cdn while loading type G order
 0
otherwise

1 i f customer demand i is met f rom cross − dock i
Fcdc
0 otherwise

1 i f cross − dock i is selected
yi
0 otherwise

1 i f input truck i precedes to input truck j
Pij
0 otherwise

qg 1 i f input truck g1 precedes to input truck g2
1 g2
0 otherwise

I f incoming truck i moves f rom supplier S
1

Xiscd to cross − dock cd while o f f loading type G order
 0
otherwise

I f incoming truck i moves f rom supplier S
1

Xisc to customer C while o f f loading type G order
 0
otherwise

I f outgoing truck j moves f rom cross − dock cd
1

j
Xcdc to customer C while o f f loading type G order
 0
otherwise

I f incoming truck i moves f rom source customer Cm
1

Xicm cn to destination customer Cn while o f f loading type G order
 0
otherwise

I f incoming truck i moves f rom source supplier Sm
1

Xism sn G to destination Supplier Sn while loading type G order
 0
otherwise

I f incoming truck i moves f rom source supplier Sm
1

Xism sn to destination Supplier Sn while o f f loading type G order
 0
otherwise

I f incoming truck i moves f rom source cross − dock cdm
1

Xicdm cdn to destination cross − dock cdn while o f f loading type G order
 0
otherwise

The first objective function minimizes the total cost of operations within the supply
chain (including transportation, fuel, delay, shortage, holding, and opening costs). The
second objective function minimizes the truck transport sequence between suppliers, cross-
docks, and customers, and the third objective function minimizes CO2 emissions from
trucks throughout the system.
Energies 2022, 15, 1530 8 of 24

" !
K S Z G
Min Z1 = ∑ ∑ ∑ ∑ (CiscdG + CisGcd ) ∗ XiscdG
i∈Kin =1 i∈s=1 i∈cd=1 i∈G=1
!
K S C G
+ ∑ ∑ ∑ CiscG ∗XiscG
∑
i∈Kin =1 i∈s=1 i∈c=1 i∈G=1
!
K S G
+ ∑ ∑ ∑ Cism sn G ∗ Xism sn G
i∈k in =1 s=1 i∈G !
K S Z G
+ ∑ ∑ ∑ ∑ (CicdsG ∗ XicdsG
i∈Kin =1 i∈s=1 i∈cd=1 i∈G=1
!
G C K
+ ∑ ∑ ∑ Cicm cn G ∗ Xicm cn G
i∈G=1 i∈c=1 i∈k in
!
K Z C G j j
+ ∑ ∑ ∑ ∑ (CcdcG ∗ XcdcG
j∈k out =1 i∈cd=1 i∈c=1 i∈G=1 ! (1)
K Z G
∑ ∑ ∑ Cicdm cdn G ∗ Xicdm cdn G
j∈Kin =1 i∈ i∈cd=1 i∈G=1
!
K C G j j
+ ∑ ∑ ∑ C cm cn G ∗ Xcm cn G
j∈Kout =1 i∈ c=1 i∈G=1
!
K C Z G j j
+ ∑ ∑ ∑ ∑ cccdG ∗XCCdG
j∈Kout =1 i∈c=1 i∈cd=1 i∈G=1
!
K C S G
+ ∑ ∑ ∑ ∑ cicsG ∗XiCsG
j∈Kin =1 i∈c=1 i∈s=1 i∈G=1
#
Z S I G K S C G
+( ∑ Oi yi + ∑ ∑ ∑ Fi ∗ XiscdG ) + [ ∑ ∑ ∑ ∑ PEkin ∗ (γi + θi ) ∗ XiscG
i∈cd=1 i∈s=1 i∈I=1 i∈G=1 i∈Kin =1 i∈s=1 i∈c=1 i∈G=1
K I C G j
+[ ∑ ∑ ∑ ∑ PEkOUT ∗ (γi + θi ) ∗ XcdcG ]
j∈Kout =1 i∈I=1 i∈c=1 i∈G=1

S C Z K K G
∑ ∑ ∑ ∑ ∑ ∑
j j j
Min Z2 = F = XiscG + XiscdG + Xism sn G + XicsG + XicdsG + XcdcG + Xcm cn G + XccdG (2)
i∈s=1 i∈c=1 i∈cd i∈kin =1 j∈kout i∈G=1

" !
S Z K
Min Z3 = [ ∑ ∑ ∑ Ckin ∗giscd ∗dscd ∗Xiscd
i∈s=1 i∈cd=1 i∈k in =1
S S G K K S C
/
+( ∑ ∑ ∑ ∑ Ckin ∗ Qkin ∗fscdG ∗WG ∗dSCdG ∗XiscdG )] + [( ∑ ∑ ∑ Ckin ∗gisc ∗dsc ∗Xisc )
i∈s=1 i∈s=1 i∈G=1 iek in i∈k in =1 i∈s=1 iec=1
S C G K K S
+( ∑ ∑ ∑ ∑ C/kin ∗ Qkin ∗fscG ∗WG ∗dSCG ∗XiscG )] + [(
∑ ∑ Ckin ∗gism sn ∗ dsm sn ∗Xism sn )
i∈s=1 i∈c=1 i∈G=1 iekin i∈kin =1 i∈s=1
K S G K Z
/ i
+( ∑ ∑ ∑ Ckin ∗ Qkin ∗fsm sn G ∗WG ∗dsm sn G ∗Xsm sn G )] + [( ∑ ∑ Ckin ∗gicdm cdn ∗dcdm cdn ∗Xicdm cdn )
i∈k in =1 i∈s=1 i∈G=1 i∈k in =1 i∈cd=1 (3)
K Z G Z C K j j
/ i
+( ∑ ∑ ∑ Ckin ∗ Qkin ∗fcdm cdn G ∗WG ∗dcdm cdn G )∗Xcdm cdn G ] + [( ∑ ∑ ∑ Ckout ∗gcdc ∗dcdc ∗Xcdc )
i∈k in =1 i∈cd=1 i∈G=1 i∈cd=1 i∈c=1 i∈k out =1
Z C G K j K C j j
/
+( ∑ ∑ ∑ ∑ Ckout ∗ Qkout ∗fcdcG ∗WG ∗dCdcG ∗XcdcG )] + [( ∑ ∑ Ckout ∗gcm cn ∗dcm cn ∗Xcm cn )
i∈cd=1 i∈c=1 i∈G=1 iek out # i∈kout =1 i∈c=1
K C G j
/
+( ∑ ∑ ∑ Ckout ∗ QkOUT ∗fcm cn G ∗WG ∗dcm cn G ∗Xcm cn G )]
i∈k out =1 i∈c=1 i∈G=1
S.t:
Energies 2022, 15, 1530 9 of 24

K C S G
∑ ∑ ∑ ∑ qGs ∗ XiscG = DcG (4)
i∈k in =1 i∈c=1 i∈s=1 i∈G=1

Z K C G j K S C G
∑ ∑ ∑ ∑ qGcd ∗X +
cdcG
∑ ∑ ∑ ∑ qGs ∗XiscG = PGs (5)
i∈cd=1 j∈k out =1 i∈C=1 i∈G=1 i∈Kin =1 i∈s=1 i∈c=1 i∈G=1

K S C G
∑ ∑ ∑ ∑ XiscG ≥ 1 (6)
i∈Kin =1 i∈s=1 i∈c=1 i∈G=1

K S G K S Z G K S C G
XicdsG = ∑ ∑ ∑ xism sn G + ∑ ∑ ∑ ∑ XiscdG + ∑ ∑ ∑ ∑ XiscG (7)
i∈Kin =1 i∈s=1 i∈G=1 i∈Kin =1 i∈s=1 i∈cd=1 i∈G=1 i∈Kin =1 i∈s=1 i∈c=1 i∈G=1

K S G K S Z G K S C G
Xism sn G = ∑ ∑ ∑ xisn sm G + ∑ ∑ ∑ ∑ Xisn cdG + ∑ ∑ ∑ ∑ XiscG (8)
i∈Kin =1 i∈s=1 i∈G=1 i∈Kin =1 i∈s=1 i∈cd=1 i∈G=1 i∈Kin =1 i∈s=1 i∈c=1 i∈G=1

K C G K C Z G
∑ ∑ ∑ ∑ ∑ ∑ ∑
j j j
XcdcG = Xcm cn G + XccdG (9)
j∈k out =1 i∈c=1 i∈G=1 i∈k out i∈c=1 i∈cd=1 i∈G=1

K C G K C Z G
Xicm cn G = ∑ ∑ ∑ Xicn cm G + ∑ ∑ ∑ ∑ XiccdG (10)
i∈Kin =1 i∈c=1 i∈G=1 i∈Kin =1 i∈c=1 i∈cd=1 i=G=1

K C G K C Z G K C Z G
∑ ∑ ∑ ∑ ∑ ∑ ∑ ∑ ∑ ∑ ∑
j j j j
Xcm cn G = Xcn cm G + XccdG + Xcdm cdn G (11)
j∈k out=1 i∈c=1 i∈G=1 j∈k out =1 i∈c=1 i∈cd=1 i∈G=1 j∈k out =1 i∈c=1 i∈cd=1 i∈G=1

K C G K C Z G
XiscG = ∑ ∑ ∑ Xicm cn G + ∑ ∑ ∑ ∑ XiccdG (12)
i∈Kin =1 i∈c=1 i∈G=1 i∈Kin =1 i∈c=1 i∈cd=1 i∈G=1

K S Z G K C Z G
∑ ∑ ∑ ∑ qGs ∗ XiscdG = ∑ ∑ ∑ ∑
j
qGcd ∗XcdcG (13)
i∈Kin =1 i∈s=1 i∈cd=1 i∈G i∈Kin =1 i∈c=1 i∈cd=1 i∈G=1

S Z G
∑ ∑ ∑ XiscdG ∗ qGs ≤ Qi (14)
i∈s=1 i∈cd=1 i∈G=1
C Z G
∑ ∑ ∑
j
XcdcG ∗ qGcd ≤ Qj (15)
i∈c=1 i∈cd=1 i∈G=1
S C G
∑ ∑ ∑ (AiscG + TiscG + ULiscG + wi ) ∗ XiscG ≤ AicsG (16)
i∈s=1 i∈C=1 i∈G=1
Z C G
∑ ∑ ∑
j j j j
(AcdcG + TcdcG + ULcdcG + wi )∗XcdcG ≤ AicsG (17)
i∈cd=1 i∈C=1 i∈G=1
K C Z G
∑ ∑ ∑ ∑
j j
XcdcG + XCCdG ≤ 1 + Fcdc (18)
j∈Kout =1 i∈c=1 i∈cd=1 i∈G=1

K S Z G
∑ ∑ ∑ ∑ XiscdG ≤ yi (19)
i∈kin=1 i∈s=1 i∈cd=1 i∈G

γi ≥ Ei − AiscG (20)
θi ≥ AiscG − Li (21)
 
j j j j
XicdsG , Xism sn G , XiscdG , XiscG , Xicm cn G , XcdcG , Xcm cn G , XiccdG , XccdG , Xcdm cdn G , yi ∈ {0, 1}, (22)
Energies 2022, 15, 1530 10 of 24

 
j j j j
Xicds , Xism sn , Xiscd , Xisc , Xicm cn , Xcdc , Xcm cn , Xiccd , Xccd , Xcdm cdn , Fcdc ∈ {0, 1} (23)

I = 1, 2, 3, . . . , I, c = 1, 2, 3, . . . , C , G = 1, 2, 3, . . . , G j = 1, 2, 3, . . . , J (24)
All variables ≥ 0, ∀ i, s, c and h (25)

2.2. Description of Objective Functions and Constraints

The first objective function minimizes the cost of transporting vehicles throughout
the system. The second objective function minimizes the transport sequence of trucks
throughout the system. The objective is utilized to sequence inbound and outbound trucks
to minimize the makespan. The third objective function minimizes the amount of carbon
dioxide emitted from vehicles throughout the system. Each route between origin and
destination has two sections. In the first part, the truck is unloaded, and in the second part,
the truck is loaded.
Constraint (1) ensures that the total number of units of type G order product loaded
from the supplier (s) in the truck (i) is precisely equal to the number of units of type (h)
order product that the customer (c) needs. Constraint (2) ensures that the total number of
order types (h) that are loaded from the supplier (s) in the entry truck (i) and move to the
customer (c), as well as the total number of order types (h) that are loaded from the cross-
docks (cd) in the output truck (j), are loaded and transferred toward the customer (c) and
are exactly equal to the sum of the number of h-type order products which are produced
by the supplier (s). Constraint (3) ensures that at least one of the incoming trucks (i) loads
the supply (s) of type G product and transports all products to the customer (c) at once.
Constraint (4) ensures that if the incoming truck (i) moves towards the supplier (s), it
selects three modes to exit the supplier. The first case is that the truck moves towards
another supplier, the second one moves towards one of the cross-docks (cd), and the last
one moves towards the customer (c). Constraint (5) ensures that if the incoming truck (i)
moves from the source supplier (Sm) to the destination supplier (Sn), it selects one of three
modes to exit the supplier (Sn). The first case is that the truck is moving towards another
supplier, the second one is moving towards one of the cross-docks (cd), and the third one is
moving towards the customer (c). Constraint (6) ensures that if the outgoing truck (j) moves
towards the customer (c), it selects one of two modes to exit the customer (c). The first case
is that the truck moves towards another customer and the second moves towards one of the
cross-docks (cd). Constraint (7) ensures that if the incoming truck (i) moves from the source
customer (cm) to the destination customer (cn), it selects one of two modes to exit the
destination customer (cn). The first case is that the truck moves towards another customer
and the second moves towards one of the cross-docks (cd). Constraint (8) is similar to
constraint (7), and the difference between these two constraints is that in constraint (8), the
outgoing truck (j) is considered in the constraint, and the third case, the outgoing truck
crosses from a cross-dock to another cross-dock. Constraint (9) ensures that if the incoming
truck (i) moves from the supplier (s) to customer (c), it selects one of two modes to exit
the customer (c). The first case is that the truck is moving towards another customer, and
the second case is that the truck is moving towards the cross-dock (cd). Constraint (10)
ensures that the number of units of product that the incoming truck (i) transports from the
supplier (s) to one of the cross-docks (cd) and unloads at the cross-dock is precisely equal
to the number of units of products in the outgoing truck (j) loaded for the customer (c).
Constraints (11) and (12) relate to the capacity of incoming and outgoing trucks and ensure
that trucks do not load products beyond their capacity. Constraint (13) ensures that the total
time of arrival of the vehicle to the customer, the time elapsed between the manufacturer
and the customer, the time of unloading the cargo at the customer’s place, and the waiting
time of the vehicle should not exceed the specified time of departure for each customer.
Constraint (14) guarantees that the total time of arrival of the vehicle to the customer, the
time elapsed between the cross-dock and the customer, the time of unloading the cargo at
the customer’s place, and the waiting time of the vehicle should not exceed the specified
time of departure of the vehicle from the customer’s location. Constraint (15) ensures that a
Energies 2022, 15, 1530 11 of 24

customer is assigned to a cross-dock if a truck connects the two. Constraint (16) ensures that
inbound truck (i) from the supplier (i) only moves to a cross-dock. Constraints (17) and (18)
are soft to calculate the number of early and late trucks to provide customer service in the
norm time window.

3. Solution Methodology
In a small example, the designed mathematical model was first solved in GAMZ
software by the є-constraint method [37]. Then, with the development of the example, due
to the large volume of model variables over time, it was not possible to solve with GAMZ
software. Therefore, it was solved using NSGA-II MOPSO meta-heuristic algorithms.

3.1. Genetic Algorithm

Charles Darwin’s theory, proposed in 1859, occupied a special place in optimization
problems [38]. This theory is based on the best evolution, and it can be considered a starting
point for evolutionary calculations. Genetic algorithms are random search algorithms
whose idea is derived from nature. Genetic algorithms for classical optimization have
successfully solved linear, convex, and similar problems, but genetic algorithms are much
more efficient for solving discrete and nonlinear problems [39]. In nature, better generations
emerge from a combination of better chromosomes. In the meantime, there are sometimes
mutations in the chromosomes that may improve the next generation. Genetic algorithms
also use this idea to solve problems [40]. Figure 2 show the schematic procedure of the
NSGA-II algorithm, and then the details of the proposed genetic algorithm are described.

Figure 2. A schematic procedure of the NSGA-II algorithm.

Energies 2022, 15, 1530 12 of 24

3.1.1. Chromosome Structure

Due to the complexity of the problem and the model, the chromosome consists of three
components. These three components include the truck’s movement between the origins
and destinations and the loading modes of various products. First, how to produce the
problem chromosome is described, and then for each example of the problem, in three levels:
small, medium, and large, according to the number of trucks, manufacturers, cross-docks,
customers, determination and capacity of trucks, time loading, unloading and handling
at the cross-dock, distances between origins and destinations, and customer demand, the
amount of production of producers are produced as a function of uniform distribution.
Table 3 shows the truck movement modes between the sources. Trucks identify
products and loading locations before starting from the cross-dock. In Table 3, for example,
four primary sources are considered: two suppliers and two cross-docks. According to
Table 3, if the truck moves to the supplier or the cross-dock, it is assigned the number
one; otherwise, it is given the number zero. The second column of Table 3, which contains
the numbers (0, 0, 0, 1), states that the truck moves only towards supplier 1 and loads
the products of supplier 1. Briefly in the table below, S1: supplier 1, S2: supplier 2, CD1:
cross-dock 1, and CD2: cross-dock 2.

Table 3. Truck motion matrix at the origins.

S1S2CD1CD2
S1S2CD1

S1S2CD2
CD1CD2
S1CD1

S1CD2

S2CD1

S2CD2
Source

S1S2
CD1

CD2
S1

S2

CD1 1 1 0 1 1 0 1 0 1 0 1 1 1
S1 0 0 0 0 0 0 0 1 1 1 0 0 0
CD2 0 0 1 0 1 1 0 1 0 0 1 1 1
S2 0 0 0 0 0 1 1 0 0 1 1 1 1

Table 4 show the trucks’ destinations after passing through the sources. For example,
in this table, four main destinations are considered, including two customers and two
cross-docks, and trucks can transport different modes of products to their destinations.
According to Table 4, if the truck moves to the customer or the intersection dock, it is
assigned the number one; otherwise, it is given the number zero. The second column of
Table 4, which contains the numbers (0, 0, 0, 1), states that the truck moves only to customer
one and unloads the customer’s products. Briefly in Table 4, C1: customer 1 and: C2:
customer 2 and CD1: cross-dock 1 and CD2: cross-dock 2.

Table 4. Truck movement matrix at destinations.

C1C2CD1CD2
Destination

C1C2CD1

C1C2CD2
C1CD1

C1CD2

C2CD1

C2CD2
C1C2
CD1

CD2
C1

C2

CD1 1 1 0 0 1 1 1 0 1 1 1 1
C1 0 0 0 0 0 0 0 1 1 0 0 0
CD2 0 0 1 0 1 1 0 1 0 1 1 1
C2 0 0 0 1 0 0 1 0 0 0 0 0

For example, four different types of products are used. These four types of products
are delivered to customers by production suppliers and trucks. Table 5 show the different
modes of loading products from suppliers in trucks. A truck can load other methods of
Energies 2022, 15, 1530 13 of 24

products in its truck. According to Table 5, if the truck loads one of the product ordering
modes, it will be number one; otherwise, it will be zero. The second column of Table 5,
which contains the numbers (0, 0, 0, 1), states that the truck only loads product A.
Table 5. Matrix of product ordering modes.

Ordering
A B C D AB AC AD BC BD CD ABC ABD ACD BCD ABCD
Mode
A 1 0 0 0 1 1 1 0 0 0 1 1 1 0 1
B 0 1 0 0 1 0 0 1 1 0 1 1 0 1 1
C 0 0 1 0 0 1 0 1 0 1 1 0 1 1 1
D 0 0 0 1 0 0 1 0 1 1 0 1 1 1 1

3.1.2. The Structure of the Cross Method

In the cross method, the production of new children is from the elite parents of the
current generation. New children should inherit the main characteristics of parents and be
more privileged than them. How children are produced depends on the structure of the
chromosome. Using the tournament method as a selection mechanism, two chromosomes
are selected as parents. The single-point cross is then performed separately for each part of
the chromosome using row (horizontal) and columnar (vertical), randomly selected and
described as follows. Table 6 show two samples of the parent chromosome to produce
new offspring.
Table 6. Two hypothetical parent chromosomes.

First Parent
1 1 0 1 1 0 1 0 1 0 1 1 1
0 0 0 0 0 0 0 1 1 1 0 0 0
0 0 1 0 1 1 0 1 0 0 1 1 1
0 0 0 0 0 1 1 0 0 1 1 1 1
Second Parent
0 0 0 1 0 0 0 1 0 0 1 1 1
0 1 0 0 0 0 0 0 1 1 0 0 0
1 0 0 0 1 1 0 1 1 0 1 0 1
0 0 1 0 1 1 1 0 0 1 1 1 1

A—Array cross operator: In this case, we produce a random number in the range
of the number of random devices. The generated random number indicates the array
location of the cross in the parent chromosomes. Then, we replace the corresponding
genes of the parents from the desired array, and new offspring are produced. For example,
array 19 is selected at the cross of the array site 19 in the parent chromosomes, and then the
corresponding genes of the parents are shifted from the array site 19 times in Table 7, and
new offspring are produced.
B—Horizontal cross operator: In this case, we generate a random number in the
range of the number of rows in the matrix at random. The generated random number
indicates the location of the cross in the parent chromosomes. The selected line then
replaces the corresponding genes of the parents, and new offspring are produced. For
example, if the second row is selected, the junction of the second row is selected on the
parent chromosomes, and then the corresponding parent genes are switched from the
second row, as shown in Table 8, and new offspring are produced.
Energies 2022, 15, 1530 14 of 24

Table 7. Children of the cross-array operator.

The First Child

0 0 0 1 0 0 0 1 1 0 1 1 1
0 1 0 0 0 0 1 1 1 1 0 0 0
1 0 0 0 1 1 1 0 0 0 1 1 1
0 0 1 0 0 1 1 0 0 1 1 1 1
The Second Child
1 1 0 1 1 0 0 1 0 0 1 1 1
0 0 0 0 0 0 0 0 1 1 0 0 0
0 0 1 0 1 1 0 1 1 0 1 0 1
0 0 0 0 1 1 1 0 0 1 1 1 1

Table 8. Children of linear (horizontal) cross operator.

The First Child

0 0 0 1 0 0 0 1 0 0 1 1 1
0 1 0 0 0 0 0 0 1 1 0 0 0
0 0 1 0 1 1 0 1 0 0 1 1 1
0 0 0 0 0 1 1 0 0 1 1 1 1
The Second Child
1 1 0 1 1 0 1 0 1 0 1 1 1
0 0 0 0 0 0 0 1 1 1 0 0 0
1 0 0 0 1 1 0 1 1 0 1 0 1
0 0 1 0 1 1 1 0 0 1 1 1 1

C—Vertical cross operator: In this case, we generate a random number in the range
of matrix columns. The generated random number indicates the location of the cross in the
parent chromosomes. The selected column then replaces the corresponding genes of the
parents, and new offspring are produced. For example, the fourth column is selected, the
intersection is performed from the location of the fourth column in the parent chromosomes,
and then the corresponding parent genes are switched from the location of the fourth
column according to Table 9, and new offspring are produced.
Table 9. Children of the linear (vertical) cross operator.

The First Child

1 1 0 1 1 0 1 0 1 0 1 1 1
0 0 0 0 0 0 0 1 1 1 0 0 0
0 0 1 0 1 1 0 1 0 0 1 1 1
0 0 0 0 0 1 1 0 0 1 1 1 1
The Second Child
0 0 0 1 1 0 1 0 1 0 1 1 1
0 0 0 0 0 0 0 1 1 1 0 0 0
1 0 0 0 1 1 0 1 0 0 1 1 1
0 0 1 0 0 1 1 0 0 1 1 1 1
Energies 2022, 15, 1530 15 of 24

3.1.3. Structure of the Mutation Method

The mutation operator is mainly used to bring about transformation diversity and
prevent divergence in the population. Due to the chromosome structure, the mutation
operator, similar to the crossover operator, is performed as one of the array methods: row
(horizontal) and columnar (vertical), which are randomly selected by the algorithm.
A—Array mutation operator: In this case, we produce a random number in the range
of random arrays. The generated random number indicates the location of the mutation
on the parent chromosome. Then, if the number in the array location is zero, it becomes
one, and if it is one, it becomes zero. For example, array 19 is selected in parent one
and the mutation is performed from array location 19 in the parent chromosomes. Then
the corresponding parent gene from array location 19 is replaced by Table 10, and a new
chromosome is produced.
Table 10. New chromosome derived from the array mutation operator.

1 1 0 1 1 0 1 0 1 0 1 1 1
0 0 0 0 0 0 0 1 1 1 0 0 0
0 0 1 0 0 1 0 1 0 0 1 1 1
0 0 0 0 0 1 1 0 0 1 1 1 1

B—Horizontal mutation operator: In this case, we generate a random number in the

range of the number of rows in the matrix. The generated random number indicates the
location of the mutation on the parent chromosome. The corresponding parent genes are
then selected from the row location, which has values of zero and one, and we move the
values of the two, and a new child is produced. For example, the second row is selected
in the first parent and the mutation is performed from the second-row location in the
parent chromosome. Then the corresponding parent genes are moved from the second-row
location in Table 11, and a new chromosome is produced.
Table 11. New chromosome derived from the horizontal mutation operator.

1 1 0 1 1 0 1 0 1 0 1 1 1
1 1 1 1 1 1 1 0 0 0 1 1 1
0 0 1 0 1 1 0 1 0 0 1 1 1
0 0 0 0 0 1 1 0 0 1 1 1 1

C—Vertical mutation operator: In this case, we generate a random number in the

range of the number of columns in the matrix. The generated random number indicates the
location of the mutation on the parent chromosome. The corresponding parent genes are
then selected from the column location, which has values of zero and one, and we move
the values of the two, and a new child is produced. For example, the seventh column is
selected in the first parent and the mutation is performed from the seventh-column location
in the parent chromosome. Then the corresponding parent genes are moved from the
seventh-column location in Table 12, and a new chromosome is produced.
Table 12. New chromosome derived from the vertical mutation operator.

1 1 0 1 1 0 0 0 1 0 1 1 1
0 0 0 0 0 0 1 1 1 1 0 0 0
0 0 1 0 1 1 1 1 0 0 1 1 1
0 0 0 0 0 1 0 0 0 1 1 1 1
Energies 2022, 15, 1530 16 of 24

3.2. Particle Swarm Optimization Algorithm—PSO

Kennedy and Eberhart (1995) presented the PSO algorithm in two papers for opti-
mization problems whose continuous nature indicates the answers [41]. The population
of answers is called a swarm. Each answer is like a bird in a group of birds and is called
a particle. It is similar to the chromosome in the genetic algorithm. All particles have a
fitness value calculated using the fitness function, and the particle fitness function must
be optimized. The velocity vector of that particle determines the direction of motion of a
particle. The PSO algorithm starts with random solutions (particles) and then seeks the
optimal solution by synchronizing the particles in each iteration. If the decision variables,
also their particles, are zero and one, each particle’s velocity and position vectors in each
iteration of the algorithm are calculated according to Equations (26)–(29).

Vit = w.Vit−1 + c1 .r1 .(pBesti − xit ) + c2 .r2 .(nBesti − xit ) (26)

− Vmax ≤ Vit ≤ Vmax (27)

si = 1/1 + eVit (28)

1 ρ ≤ si
xit = (29)
0 otherwise
According to the relation of 19, new velocity vectors of each particle based on the
previous velocity of the particle itself (Vit−1 ), the best position the particle has ever reached
(pBesti ) and the position of the best particle in the neighborhood of the particles obtained
so far (nBesti ), are calculated. If the neighborhood of each particle contains all the particles
in the group, then nBesti indicates the position of the best particle in the group, denoted
by gBest. r1 and r2 are two random numbers with a uniform distribution between [0, 1]
generated independently of each other. c1 and c2 are referred to as learning coefficients,
they control the effect of pBest and nBest during the search process. w represents the
weight coefficient of inertia. The value limits the particle velocity vector (Vmax ). Vmax is a
constraint that controls the global search capability of a particle group. Using the relation of
20 vectors, the velocity of each particle is converted to the change probability vector. In the
above relation, si indicates the probability that xit is equal to 1. Then, using the relation of
22 vectors, the position of each particle is updated. In the above relation, ρ is random with
a uniform distribution between zero and one. The pseudo-code of group PSO is shown
in Table 13.

Table 13. PSO pseudo-code.

For each particle
Initialize particle
End For
Do
For each particle
Calculate fitness value of the particle fp
/*updating particle’s best fitness value so far)*/
If fp is better than pBest
set current value as the new pBest
End For
/*updating population’s best fitness value so far)*/
Set gBest to the best fitness value of all particles
For each particle
Calculate particle velocity according to equation
Update particle position according to equation
End For While maximum iterations OR minimum error criteria is not attained

3.3. Validation of the Designed Model

Although many articles studied the issue of truck scheduling and transportation time
minimization at the intersection dock, the issue of cost minimization (transportation, load-
Energies 2022, 15, 1530 17 of 24

ing, unloading, and relocation) throughout the supply chain, considering the relationships
and assumptions, was not previously studied. This article represents a new beginning for
future work in this field. Therefore, since it is impossible to compare the results with other
existing articles, a numerical example is presented in the next section, and the answers to
the problem are compared using genetic algorithms and particle swarm.

Method of Generating Random Problems

Sample production problems are randomly generated in small, medium, and large
groups. From each dimension of 7 sample problems, 21 sample problems are described
according to Table 14 of the production. Then, the computational results will be presented,
and finally, by repeating the algorithm implementations and using the uniform distribution
function, the problem parameters are generated, and new numbers are generated in each
sample. Additionally, the sensitivity analysis of the results is analyzed using the number
of trucks, and in the end, the two algorithms are compared using different performance
criteria. The solution of the model was obtained using MATLAB (R2019a) software. The
problem was that the trucks would move from suppliers to customers or cross-dock and
unload the loaded products after loading the products. To better understand how the two
algorithms work, the problem-solving details of Example 6 are shown in a small size, which
is almost the most straightforward example problem.
Table 14. Fixed parameters of the problems.

Problem 1 Problem 2 Problem 3

Parameters
1 2 3 4 5 6 7 1 2 3 4 5 6 7 1 2 3 4 5 6 7
Number of
2 4 4 3 4 3 3 3 4 2 4 4 3 3 12 11 13 14 11 12 11
incoming trucks
Number of
2 4 3 4 4 3 2 2 3 3 2 3 2 2 11 12 11 12 11 10 12
outgoing trucks
Number of suppliers 4 3 4 4 5 3 2 10 9 9 8 7 8 11 18 17 17 19 18 16 20
Number of cross-dock 4 3 3 7 4 3 3 8 9 9 7 8 6 6 12 10 11 12 10 10 14
Product types 10 6 8 10 9 10 11 12 12 14 12 11 13 15 20 16 18 20 20 20 25
Number of customer 5 3 4 8 4 3 3 11 12 10 9 10 11 15 22 20 23 21 20 18 20

Several parameter values are generated using the uniform distribution function in this
problem. Parameters that are generated randomly include the capacity of the trucks, the
amount of customer demand, the amount of production of the manufacturers, the number
of products, the fuel consumption rate, and the distance traveled by the trucks between
the origins and the destinations. Sample parameter values are created and given in the
relevant tables for each problem. Table 15 list the minimum and maximum values that the
parameters can be assigned.
Table 15. Random parameters of the problem.

Customer demand U(0,30) U(0,30) U(0,30)

Product supply rate U(0,30) U(0,30) U(0,30)
Product weight U(0,10) U(0,10) U(0,10)
Incoming truck capacity U(0,10) × 1000 U(0,10) × 1000 U(0,10) × 1000
Outgoing trucks capacity U(0,10) × 1000 U(0,10) × 1000 U(0,10) × 1000
Origin between destination distance U(1,100) U(1,100) U(1,100)
Number of product type G U(1,10) U(1,10) U(1,10)
Fuel consumption rate U(1,20) U(1,20) U(1,20)

Two meta-heuristic algorithms for solving multi-objective problems, including the

NSGA-II faulty sorting genetic algorithm and the MOPSO multi-objective particle swarm
algorithm, were used to solve the proposed model. The Taguchi method was used to
Energies 2022, 15, 1530 18 of 24

adjust the parameters of the algorithms. The advantage of the Taguchi method over
other experimental design methods is that in addition to cost, obtaining optimal levels of
parameters can be achieved in less time. One of the most important steps in this method is
to select an orthogonal array that estimates the effects of factors on the mean response and
variation. In this paper, the most suitable three-level test design is identified, and according
to the Taguchi standard orthogonal arrays, the L9 array is selected as the appropriate test
design to parameterize the proposed algorithms. A statistical measure of performance is
considered the S/N (signal to noise) ratio to adjust the optimal parameters, which includes
the average and changes. This ratio is desirable. The considered response variable is the
average of the four standard indices MID (mean ideal distance), MD (maximum spread or
diversity), spacing, and NPS (number of pareto solution) for multi-objective meta-heuristic
algorithms. Since this response variable is less, its corresponding S/N ratio is considered
as Equation (30). The proposed meta-heuristic algorithms were implemented for each
Taguchi experiment, and then the S/N ratios were calculated using Minitab 18 software.
The optimal values of the parameters of each algorithm are shown in Table 16.

sum y2


S/NRatio = −10log( ) (30)
n

Table 16. Optimal values of parameters in algorithms.

Algorithm Parameters Parameter Domin Amounts

Iteration 100–300 300
Population size 50–100 100
NSGA-II
Intersection rate 0.6–0.8 0.8
Mutation rate 0.1–0.2 0.2
Iteration 100–300 300
Population size 50–100 100
MOPSO
Cognitive constants, C1 1–3 3
Social constant, C2 1–2 2

4. Results
Sample problem 6 was solved at the small size level with the fixed parameters listed in
Table 4 and the random parameters listed in Table 6 with the four algorithms NSGA-II and
MOPSO. Based on this, the sequence of truck transportation operations in sample case 6,
for all four algorithms, is presented as the output analysis of the variables according to
the following tables. Table 17 show the transport sequences of the first truck in sample
problem 6 by solving the NSGA-II algorithm, and Table 18 show the transport sequences of
the first truck in sample problem 6 using the MOPSO solution method.
Table 17. Truck transport sequence 1 with NSGA-II algorithm for sample problem 1.

Number of Shipping
Origin Destination Product Type
Products Sequence
Supplier 1 Cross-dock 1 E 7
1
Supplier 2 Cross-dock 1 F 15
Supplier 3 Cross-dock 3 G 6 2
Supplier 3 Customer 1 D 14
Supplier 3 Customer 2 D 14
Supplier 2 Customer 3 D 13
Supplier 1 Customer 3 E 17
Supplier 2 Customer 1 F 8 3
Supplier 2 Customer 2 F 16
Supplier 2 Customer 3 F 18
Energies 2022, 15, 1530 19 of 24

Table 17. Cont.

Number of Shipping
Origin Destination Product Type
Products Sequence
Supplier 2 Customer 2 G 7
Supplier 3 Supplier 1 E 7
Supplier 2 Supplier 1 F 15
Supplier 3 Supplier 1 F 16 4
Cross-dock 3 Cross-dock 1 D 14
5
Cross-dock 2 Cross-dock 1 F 16
Customer 3 Customer 1 D 14
Customer 3 Customer 2 E 8
Customer 3 Customer 2 F 18
Customer 3 Customer 2 G 13 6
Cross-dock 1 Customer 1 B 9
Cross-dock 1 Customer 3 D 14
Cross-dock 1 Customer 1 G 10
Cross-dock 1 Customer 2 G 7
Cross-dock 2 Customer 3 C 11 7
Cross-dock 3 Customer 1 E 17

Table 18. Truck transport sequence 1 with MOPSO algorithm for sample problem 1 (small level).

Number of Shipping
Origin Destination Product Type
Products Sequence
Supplier 2 Cross-dock 2 D 8
Supplier 2 Cross-dock 2 E 8
Supplier 2 Cross-dock 2 H 11 1
Supplier 2 Customer 1 D 14
Supplier 2 Customer 2 D 13
Supplier 2 Customer 1 E 17
Supplier 2 Customer 2 E 4
2
Supplier 2 Customer 3 E 8
Supplier 1 Customer 3 D 14
Supplier 2 Supplier 1 D 8
Supplier 2 Supplier 1 H 11 3
Cross-dock 2 Cross-dock 1 D 12 4
Customer 3 Customer 2 E 8
Customer 1 Customer 2 H 4 5
Cross-dock 2 Customer 3 C 17
Cross-dock 2 Customer 1 E 17
Cross-dock 2 Customer 2 E 4
Cross-dock 2 Customer 3 G 7
Cross-dock 1 Customer 3 E 17 6
Cross-dock 3 Customer 2 E 8

4.1. Comparison of NSGA-II and MOPSO Algorithms

Tables 19 and 20 show the values related to the criteria of several answers, time to
solve problems, distance from the ideal point, distance, and maximum coverage to compare
NSGA-II and MOPSO algorithms.
Execution time—One of the crucial criteria for measuring the performance quality of
an algorithm is its execution time, which in some articles is also referred to as execution
speed. This criterion becomes essential when the dimensions and complexity of the problem
increase. The execution time of these four methods in the three levels of small, medium,
and large sample problems are given in Tables 19 and 20 and Figure 3. In general, the
NSGA-II algorithm is more acceptable in terms of average runtime at both medium and
large sample problem levels.
Energies 2022, 15, 1530 20 of 24

Table 19. Criteria values of the NSGA-II algorithm.

NSGA-II
Example Size
NPS Time (S) MID DM Spacing
1 Small 100 878.22 1.0918 50964 0.8515
2 Small 97 526.19 1.0601 3351 0.9425
3 Small 100 898.23 1.0361 5745.4 0.9863
4 Small 100 1204.84 1.1609 5082.3 0.8294
5 Small 100 911.26 1.0537 5281.4 0.9158
6 Small 99 392.67 1.0077 3667.4 0.9317
7 Small 99 711.97 1.1094 4616 1.0001
Mean 99.28 798.054 1.0742 4691.4 0.9226
1 Middle 100 2011.51 1.0466 9406.5 0.8803
2 Middle 100 2213.24 1.0494 9415 0.9962
3 Middle 100 3415.81 1.0453 11,761 1.0038
4 Middle 100 4313.86 1.043 11,291 1.0075
5 Middle 100 4201.49 1.0371 12,580 1.0718
6 Middle 100 3919.35 1.0394 13,449 0.9942
7 Middle 99 5077.08 1.0556 12,997 0.963
Mean 99.85 3593.19 1.045 11,557.07 0.9881
1 Large 100 19,202.23 1.0627 19,654 1.1429
2 Large 99 21,020.54 1.0748 21,012 1.1436
3 Large 100 210,695.14 1.0452 20,654.12 1.1259
4 Large 100 230,458.87 1.0872 23,012.45 1.1248
5 Large 100 26,748.65 1.0925 24,896.87 1.0258
6 Large 99 25,874.96 1.0745 23,968.97 1.0387
7 Large 100 27,987.56 1.0998 25,984.23 1.0587
Mean 99.714 22,715.28 1.0766 22,624.79 1.0941

Table 20. Criteria values of the MOPSO algorithm.

MOPSO
Example Size
NPS Time (S) MID DM Spacing
1 Small 100 817.45 1.1166 2708.6 0.8495
2 Small 77 473.64 1.1234 1365.3 0.803
3 Small 99 831.9 1.0548 2972.3 0.9092
4 Small 94 1161.45 1.2043 2897.9 0.7604
5 Small 95 853.09 1.0943 3108.5 0.8189
6 Small 98 381.79 1.0829 2430 0.907
7 Small 95 650.25 1.1492 2324.6 0.7806
Mean 94 738.51 1.1179 2453.88 0.8326
1 Middle 100 2044.69 1.0468 6090.3 0.9312
2 Middle 92 2320.03 1.0455 7155.5 1.0476
3 Middle 100 3451.65 1.0547 6141 0.8661
4 Middle 98 4364.47 1.049 7536.4 1.0063
5 Middle 98 4113.73 1.0404 7265.8 0.9996
6 Middle 100 3996.17 1.0364 8293.4 0.9286
7 Middle 96 5258.46 1.0625 6351.8 0.8655
Mean 97.714 3649.88 1.0479 6976.3 0.9493
1 Large 100 19,795.71 1.0877 10395 0.8874
2 Large 100 22,101.36 1.0689 11,256 0.8658
3 Large 100 20,985.87 1.0589 10,365.85 0.8953
4 Large 100 21,895.3 1.0489 12,365.35 0.9587
5 Large 96 27,014.32 1.0845 12,985.36 0.8596
6 Large 100 26,579.89 1.0895 12,645.84 0.9741
7 Large 93 28,012.56 1.0114 13,586.96 0.8254
Mean 98.42 23,769.27 1.06 11,942.91 0.8951
Energies 2022, 15, 1530 21 of 24

Figure 3. Comparison between the average execution time of NSGA-II and MOPSO.

4.1.1. Distance from the Ideal Point

The MID index is used to calculate the mean distance of Pareto solutions from the
origin’s so-called ideal point. Considering the MID, it is clear that the lower the value of
this index, the higher the efficiency of this algorithm. The ideal point distance criterion,
which measures the optimal close Pareto proximity measure, was calculated for 21 sample
problems for the NSGA-II and MOPSO algorithms. In these 21 problems, the average
values of the NSGA-II algorithm in small and medium problems were less than for the
MOPSO algorithm. The DM index shows the variety of Pareto solutions. The higher the
DM value is, the better the algorithm performs.

4.1.2. Spacing
The spacing criterion for each sample problem was calculated for the NSGA-II MOPSO
algorithms, and the comparison of the mean of these two values for the sample problems
was performed at three levels to determine the most desirable algorithm. The spacing
criterion in the MOPSO algorithm was smaller than the NSGA-II algorithm and had
better performance.

4.1.3. The Most Expansion

The maximum expansion criterion for the 21 sample problems was calculated for the
NSGA-II and MOPSO algorithms. The criterion measures the maximum expansion density
of Pareto archive solutions, and a simple method was used to calculate it, i.e., without
considering the intermediate solutions, only the boundary solutions were used to plot the
extent of Pareto archive solutions. In this criterion, the average of the calculations shows that
the NSGA-II algorithm offers more comprehensive answers than the MOPSO algorithm.

5. Conclusions and Suggestions

The problem studied in this article was used to provide a model for the optimal
sequence of trucks and the cost of operations within the supply chain concerning three
objectives (minimizing the cost of transportation, minimizing the sequence of transporting
trucks, and minimizing carbon dioxide emissions) [42–45]. The mathematical model
approach is complicated due to the high number of variables and limitations related
to the number of trucks sending and receiving and the number of products required to
solve the problem. When increasing these variables, the dimensions of the problem and
Energies 2022, 15, 1530 22 of 24

the solution time increase exponentially [46,47]. Therefore, two meta-heuristic algorithms
NSGA-II and MOPSO, were used to solve the model. The results of the calculations were
compared using four criteria. On average, the NSGA-II algorithm is more acceptable in
runtime for 21 sample problems at the three small, medium, and large levels. In terms of
distance from the ideal point, the NSGA-II algorithm is better. In the spacing criterion,
which measures the density of Pareto archive solutions, the MOPSO algorithm scores better
than the NSGA-II algorithm. The NSGA-II algorithm offers broader answers than the
MOPSO algorithm at the maximum expansion criterion. Because research in this area is in
its infancy, there is much opportunity for future research. Several areas in which research
can be expanded are listed below.
Generalization of the proposed mathematical model for modelling in the presence of
cross-dock systems with many entrances and exit doors.
It was assumed that temporary storage has unlimited capacity, while temporary stor-
age is usually limited in practice. This real-world assumption can be added to the model.
Generalization of the proposed mathematical model by considering the multi-
period mode.
Generalization of the proposed mathematical model of the problem by considering
uncertain parameters (for example, considering fuzzy demand for customers) given the
complex nature of docking systems.

Author Contributions: Conceptualization, I.M.-K. and N.S.; methodology, S.G. and N.S.; software,
F.Y.; validation, I.M.-K.; formal analysis, S.G. and F.Y.; investigation, I.M.-K. and R.Č.; resources, S.G.
and R.Č.; writing—original draft preparation, S.G. and N.S.; writing—review and editing, I.M.-K.;
visualization, S.G. and R.Č.; supervision, I.M.-K. All authors have read and agreed to the published
version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: The data of this study are available from the authors upon request.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Rijal, A.; Bijvank, M.; de Koster, R. Integrated scheduling and assignment of trucks at unit-load cross-dock terminals with mixed
service mode dock doors. Eur. J. Oper. Res. 2019, 278, 752–771. [CrossRef]
2. Serrano, C.; Delorme, X.; Dolgui, A. Scheduling of truck arrivals, truck departures and shop-floor operation in a cross-dock
platform, based on trucks loading plans. Int. J. Prod. Econ. 2017, 194, 102–112. [CrossRef]
3. Seyedi, I.; Hamedi, M.; Tavakkoli-Moghadaam, R. Developing a mathematical model for a multi-door cross-dock scheduling
problem with human factors: A modified imperialist competitive algorithm. J. Ind. Eng. Manag. Stud. 2021, 8, 180–201.
4. Motaghedi-Larijani, A.; Aminnayeri, M. Optimizing the admission time of outbound trucks entering a cross-dock with uniform
arrival time by considering a queuing model. Eng. Optim. 2017, 49, 466–480. [CrossRef]
5. Gelareh, S.; Glover, F.; Guemri, O.; Hanafi, S.; Nduwayo, P.; Todosijević, R. A comparative study of formulations for a cross-dock
door assignment problem. Omega 2020, 91, 102015. [CrossRef]
6. Xi, X.; Changchun, L.; Yuan, W.; Hay, L.L. Two-stage conflict robust optimization models for cross-dock truck scheduling problem
under uncertainty. Transp. Res. Part E Logist. Transp. Rev. 2020, 144, 102123. [CrossRef]
7. Sahebi, I.G.; Mosayebi, A.; Masoomi, B.; Marandi, F. Modeling the enablers for blockchain technology adoption in renewable
energy supply chain. Technol. Soc. 2022, 101871. [CrossRef]
8. Nassief, W. Cross-dock Door Assignments: Models, Algorithms and Extensions. Ph.D. Thesis, Concordia University, Montréal, QC,
Canada, 2017. Available online: https://spectrum.library.concordia.ca/id/eprint/982540/1/Nassief_PhD_S2017.pdf (accessed
on 19 November 2021).
9. Nasiri, M.M.; Rahbari, A.; Werner, F.; Karimi, R. Incorporating supplier selection and order allocation into the vehicle routing and
multi-cross-dock scheduling problem. Int. J. Prod. Res. 2018, 56, 6527–6552. [CrossRef]
10. Chargui, T.; Bekrar, A.; Reghioui, M.; Trentesaux, D. Simulation for Pi-hub cross-docking robustness. In Service Orientation in
Holonic and Multi-Agent Manufacturing; Springer: Berlin, Germany, 2018; pp. 317–328.
11. Tirkolaee, E.B.; Goli, A.; Faridnia, A.; Soltani, M.; Weber, G.-W. Multi-objective optimization for the reliable pollution-routing
problem with cross-dock selection using Pareto-based algorithms. J. Clean. Prod. 2020, 276, 122927. [CrossRef]
Energies 2022, 15, 1530 23 of 24

12. Wang, H.; Alidaee, B. The multi-floor cross-dock door assignment problem: Rising challenges for the new trend in logistics
industry. Transp. Res. Part E Logist. Transp. Rev. 2019, 132, 30–47. [CrossRef]
13. Dondo, R.; Cerdá, J. The heterogeneous vehicle routing and truck scheduling problem in a multi-door cross-dock system. Comput.
Chem. Eng. 2015, 76, 42–62. [CrossRef]
14. Zuluaga, J.P.S.; Thiell, M.; Perales, R.C. Reverse cross-docking. Omega 2017, 66, 48–57. [CrossRef]
15. Mohtashami, A.; Tavana, M.; Santos-Arteaga, F.J.; Fallahian-Najafabadi, A. A novel multi-objective meta-heuristic model for
solving cross-docking scheduling problems. Appl. Soft Comput. 2015, 31, 30–47. [CrossRef]
16. Wisittipanich, W.; Hengmeechai, P. Truck scheduling in multi-door cross docking terminal by modified particle swarm optimiza-
tion. Comput. Ind. Eng. 2017, 113, 793–802. [CrossRef]
17. Mohtashami, A. A novel dynamic genetic algorithm-based method for vehicle scheduling in cross docking systems with frequent
unloading operation. Comput. Ind. Eng. 2015, 90, 221–240. [CrossRef]
18. Ponboon, S.; Qureshi, A.G.; Taniguchi, E. Evaluation of cost structure and impact of parameters in location-routing problem with
time windows. Transp. Res. Procedia 2016, 12, 213–226. [CrossRef]
19. Gomes, C.F.S.; Ribeiro, P.C.C.; de Matos Freire, K.A. Bibliometric research in Warehouse Management System from 2006 to 2016. In
Proceedings of the World Multi-Conference on Systemics, Cybernetics and Informatics, Orlando, FL, USA, 8 July 2018; Volume 22,
pp. 200–204.
20. Birim, Ş. Vehicle routing problem with cross docking: A simulated annealing approach. Procedia Social Behav. Sci. 2016, 235,
149–158. [CrossRef]
21. Yin, P.-Y.; Chuang, Y.-L. Adaptive memory artificial bee colony algorithm for green vehicle routing with cross-docking. Appl.
Math. Model. 2016, 40, 9302–9315. [CrossRef]
22. Mohtashami, A.; Fallahian-Najafabadi, A. Scheduling trucks transportation in supply chain regarding cross docking using
meta-heuristic algorithms. Ind. Manag. Stud. 2014, 11, 55–84.
23. Sung, C.S.; Song, S.H. Integrated service network design for a cross-docking supply chain network. J. Oper. Res. Soc. 2003, 54,
1283–1295. [CrossRef]
24. Chen, P.; Guo, Y.; Lim, A.; Rodrigues, B. Multiple crossdocks with inventory and time windows. Comput. Oper. Res. 2006, 33,
43–63. [CrossRef]
25. Sung, C.S.; Yang, W. An exact algorithm for a cross-docking supply chain network design problem. J. Oper. Res. Soc. 2008, 59,
119–136. [CrossRef]
26. Wen, M.; Larsen, J.; Clausen, J.; Cordeau, J.-F.; Laporte, G. Vehicle routing with cross-docking. J. Oper. Res. Soc. 2009, 60,
1708–1718. [CrossRef]
27. Musa, R.; Arnaout, J.-P.; Jung, H. Ant colony optimization algorithm to solve for the transportation problem of cross-docking
network. Comput. Ind. Eng. 2010, 59, 85–92. [CrossRef]
28. Dondo, R.; Méndez, C.A.; Cerdá, J. The multi-echelon vehicle routing problem with cross docking in supply chain management.
Comput. Chem. Eng. 2011, 35, 3002–3024. [CrossRef]
29. Santos, F.A.; Mateus, G.R.; Da Cunha, A.S. The pickup and delivery problem with cross-docking. Comput. Oper. Res. 2013, 40,
1085–1093. [CrossRef]
30. Morais, V.W.C.; Mateus, G.R.; Noronha, T.F. Iterated local search heuristics for the vehicle routing problem with cross-docking.
Expert Syst. Appl. 2014, 41, 7495–7506. [CrossRef]
31. Vincent, F.Y.; Jewpanya, P.; Redi, A.A.N.P. Open vehicle routing problem with cross-docking. Comput. Ind. Eng. 2016, 94, 6–17.
32. Wang, J.; Jagannathan, A.K.R.; Zuo, X.; Murray, C.C. Two-layer simulated annealing and tabu search heuristics for a vehicle
routing problem with cross docks and split deliveries. Comput. Ind. Eng. 2017, 112, 84–98. [CrossRef]
33. Rahbari, A.; Nasiri, M.M.; Werner, F.; Musavi, M.; Jolai, F. The vehicle routing and scheduling problem with cross-docking for
perishable products under uncertainty: Two robust bi-objective models. Appl. Math. Model. 2019, 70, 605–625. [CrossRef]
34. Baniamerian, A.; Bashiri, M.; Tavakkoli-Moghaddam, R. Modified variable neighborhood search and genetic algorithm for
profitable heterogeneous vehicle routing problem with cross-docking. Appl. Soft Comput. 2019, 75, 441–460. [CrossRef]
35. Zhou, B.; Zong, S. Adaptive memory red deer algorithm for cross-dock truck scheduling with products time window. Eng. Comput.
2021, 38, 3254–3289. [CrossRef]
36. Liao, T.W. Integrated Outbound Vehicle Routing and Scheduling Problem at a Multi-Door Cross-Dock Terminal. IEEE Trans.
Intell. Transp. Syst. 2020, 22, 5599–5612. [CrossRef]
37. Safari, H.; Ajali, M.; Ghasemiyan Sahebi, I. Determining the strategic position of an educational institution in the organizational life cycle
with fuzzy approach (Case Study: Social Sciences Faculty of Khalij Fars University). Mod. Res. Decis. Mak. 2016, 1, 117–138.
38. Khalili-Damghani, K.; Tavana, M.; Santos-Arteaga, F.J.; Ghanbarzad-Dashti, M. A customized genetic algorithm for solving
multi-period cross-dock truck scheduling problems. Measurement 2017, 108, 101–118. [CrossRef]
39. Corsten, H.; Becker, F.; Salewski, H. Integrating truck and workforce scheduling in a cross-dock: Analysis of different workforce
coordination policies. J. Bus. Econ. 2020, 90, 207–237. [CrossRef]
40. Maknoon, Y.; Laporte, G. Vehicle routing with cross-dock selection. Comput. Oper. Res. 2017, 77, 254–266. [CrossRef]
41. Eberhart, R.; Kennedy, J. A new optimizer using particle swarm theory. In Proceedings of the MHS’95. Proceedings of the Sixth
International Symposium on Micro Machine and Human Science, Nagoya, Japan, 4–6 October 1995; pp. 39–43.
Energies 2022, 15, 1530 24 of 24

42. Sahebi, I.; Masoomi, B.; Ghorbani, S.; Uslu, T. Scenario-based designing of closed-loop supply chain with uncertainty in returned
products. Decis. Sci. Lett. 2019, 8, 505–518. [CrossRef]
43. Kusolpuchong, S.; Chusap, K.; Alhawari, O.; Suer, G. A Genetic Algorithm Approach for Multi Objective Cross Dock Scheduling
in Supply Chains. Procedia Manuf. 2019, 39, 1139–1148. [CrossRef]
44. Arab, A.; Sahebi, I.G.; Modarresi, M.; Ajalli, M. A Grey DEMATEL approach for ranking the KSFs of environmental management
system implementation (ISO 14001). Calitatea 2017, 18, 115.
45. Arab, A.; Sahebi, I.G.; Alavi, S.A. Assessing the key success factors of knowledge management adoption in supply chain. Int. J.
Acad. Res. Bus. Soc. Sci. 2017, 7, 2222–6990. [CrossRef]
46. Sayed, S.I.; Contreras, I.; Diaz, J.A.; Luna, D.E. Integrated cross-dock door assignment and truck scheduling with handling times.
Top 2020, 28, 705–727. [CrossRef]
47. Heidari, F.; Zegordi, S.H.; Tavakkoli-Moghaddam, R. Modeling truck scheduling problem at a cross-dock facility through a
bi-objective bi-level optimization approach. J. Intell. Manuf. 2018, 29, 1155–1170. [CrossRef]