Sustainable Operations and Computers 2 (2021) 64–70

Contents lists available at ScienceDirect

Sustainable Operations and Computers

journal homepage:
http://www.keaipublishing.com/en/journals/sustainable-operations-and-computers/

A green multi-period inventory routing problem with pickup and split

delivery: A case study in flour industry
Meysam Mahjoob a, Seyed Sajjad Fazeli b,∗, Leyla Sadat Tavassoli c, Mirpouya Mirmozaffari c,
Soodabeh Milanlouei d
a
Department of Industrial Engineering, University of Tehran, Fooman, Rasht, Iran
b
Department of Industrial and System Engineering, Wayne State university, Detroit, MI, USA
c
Department of Industrial Manufacturing and Systems Engineering, University of Texas at Arlington, Arlington, TX USA
d
Center for Complex Network Research, Northeastern University, Boston, MA, USA

a r t i c l e i n f o a b s t r a c t

Keywords: In this study, we proposes a multi-period inventory routing problem with pickup and split delivery for distributing
Inventory routing problem flour among bakeries. This study aims to determine the routes for heterogeneous vehicles with different loading
Pickup and split delivery capacity and, at the same time, decrease the cost of the system by enabling the pickup and split delivery at
Vendor managed inventory
customer nodes. Also, we incorporate the environmental aspects into our modeling framework to reduce the
Supply chain management
adverse effect of transportation operations. We formulate the problem as a bi-objective multi-period inventory
Pareto solution
routing problem with pickup and split delivery. The proposed framework incorporates different strategical and
operational decisions to optimize the supply chain network. We evaluate the performance of the model with and
without the pickup option through an illustrative example. From the algorithm perspective, we implemented
a priori method to solve the proposed formulation. We performed numerous experiments on real instances to
showcase the application of the proposed framework on real-world problems. Finally, we provided the Pareto
solution set for decision-makers to choose the best solutions based on their preference.

1. Introduction an integrated inventory routing problem (IRP) [4]. IRP addresses the
vehicle routing problem simultaneously integrated with the inventory
Over the past few decades, information technology availability has management to achieve a better overall performance in a supply chain
resulted in the development of new businesses in supply chain manage- system. Generally, IRP can be described as follows: A supply chain sys-
ment, such as Vendor Managed Inventory (VMI) [1]. VMI represents a tem consisting of a supplier and some customers with specific demands
paradigm where the vendor (supplier) is responsible for managing cus- dispersed in a particular geographic area. Products are shipped from
tomers’ (retailers) inventory through day-to-day monitoring and deci- the supplier to the customers by a fleet of vehicles in a given time hori-
sion making for customer replenishment policies. In a VMI system, un- zon. Each customer is visited by some vehicles (split delivery) in each
like traditional methods where customers monitor their inventory and period. The supplier determines the replenishment policy for each cus-
decide on the time and the amount of the product to be ordered, a ven- tomer as well as the routes for each vehicle, ensuring that there will be
dor controls and manages the level of its customers’ inventory. In this no shortage. IRP’s objective is to minimize overall logistics costs while
type of distribution system, the decision on the amount of the inventory meeting the customers’ demand within a limited planning horizon so
to be held by the supplier and the retailer is affected by the delivery time that any shortages for each customer be avoided [5,6]. There are three
and delivery amount of the retailers’ and vehicles’ capacity [2,3]. There- types of decisions that need to be made: (1) When each customer should
fore, such systems need to employ simultaneous decision making to be be visited; (2) how much should be delivered to each customer [7,8];
cost-effective. There are many potential benefits of using a VMI system. (3) What routes should be selected for every vehicle (Vehicle Routing
For example, a vendor can increase the service level and reduce distri- Problem (VRP)) [9–13]. Over the past decades, many researchers have
bution costs through efficient use of the vehicles. On the other hand, studied various extensions of the VRP. Few studies focused on the lit-
customers can spend fewer resources to manage their inventory in the erature review of the inventory-routing problem. A research by Feder-
ordering process. So, the vendor derived from a VMI system must solve gruen and Simchi-Levi [9] presented one of the earliest works, where

∗
Corresponding author.
E-mail addresses: [email protected] (M. Mahjoob), [email protected] (S.S. Fazeli), [email protected] (L.S. Tavassoli),
Mirpouya.mirmozaff[email protected] (M. Mirmozaffari), [email protected] (S. Milanlouei).

https://doi.org/10.1016/j.susoc.2021.04.002
Received 25 December 2020; Received in revised form 26 March 2021; Accepted 5 April 2021
Available online 20 April 2021
2666-4127/© 2021 The Author(s). Published by Elsevier B.V. on behalf of KeAi Communications Co., Ltd. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
M. Mahjoob, S.S. Fazeli, L.S. Tavassoli et al. Sustainable Operations and Computers 2 (2021) 64–70

they analyzed the vehicle routing and inventory-routing problems by ing the proposed model and then introduces the solution methodology
incorporating the bin-packing features into the VRP. Authors in [14] re- that we implemented to solve real-world instances. Section 4 presents
viewed the integrated analyses of the production-distribution systems. the extensive computational experiments. Finally, Section 5 provides
They focused on the logistics aspect of inventory routing problems. A concluding remarks.
comprehensive review was presented by Andersson et al. [15] with an
emphasis on the industrial aspects of the inventory routing problem. In
general, the inventory-routing problem is of interest in a wide range 2. Model formulation
of industries, including the gas industry [16,17], chemical and petro-
chemical products [18,19], automobile [20] and foods [21,22]. Vehicle 2.1. Problem definition
Routing Problem with Pickup and Delivery (VRPPD) is a generalized
version of the VRP that not only includes deliveries but also enables The GMIRPSD consists of a distributor (supplier) and some customers
pickup of products from the customers. VRPPD can consist of differ- (retailer) with certain demands dispersed in a geographic area. Prod-
ent variants such as first delivery and then pickup, mixed delivery and ucts are shipped from the distributor by a fleet of vehicles with differ-
pickup, and simultaneous delivery and pickup. Researchers in [23] clas- ent capacities in a given planning horizon. Customers’ storage capacity
sified pickup and delivery problems (PDPs) into three different groups. and storage costs are considered to be different. Therefore, when a cus-
The first group consists of many-to-many problems, in which any node tomer’s storage cost is high, or the capacity is not enough to store prod-
can serve as a source or as a destination for any product. Products can ucts for the upcoming period, we can use the storage space of another
be delivered from one of the nodes to other nodes. Problems in the sec- customer that has available storage capacity or lower storage cost. This
ond group are called one-to-many-to-one problems. In these problems, can reduce the storage cost and consequently reduce the total cost of the
products are distributed from depots to customers and also from cus- system. The GMIRPSD model seeks to determine vehicle routes and the
tomers to depots versa. The third group is one-to-one problems where quantity received by and delivered to each customer during the planning
each product has a given origin and a given destination [24]. Further- horizon. The problem’s objective is to minimize system’s costs, includ-
more, the work in [25] considered a pickup and delivery inventory prob- ing inventory cost, transportation cost, delivery, and receiving costs, as
lem with time window in a closed-loop supply chain. They developed well as reducing carbon CO2 emissions caused by transportation opera-
a mixed integer programming model containing transportation, inven- tions. Transportation cost consists of two components: the cost related
tory, production cost as well as a penalty cost associated with the time to the distance traveled and the vehicle’s fixed cost.
length of routes. The authors in [26] formulated a multi-period inven-
tory routing problem with pickup and delivery for a case study of the • Vehicles are heterogeneous, meaning they have different load capac-
replenishment of ATMs in Netherlands. To overcome the complexity of ities.
the model, they decomposed the problem into several sub-problems and • The travel cost between the distributor and customers is determinis-
then solved them using a branch and cut approach. Since VRPPD is an tic, symmetrical, and fixed in each period during the planning hori-
extension of VRP, it is Np-Hard. During the past decades, authors de- zon.
veloped different approaches to overcome the complexity of the VRPPD • The demand of each customer is deterministic and definite in each
and its different variants which can broadly categorized into heuristics period.
& meta-heuristics [27–30], and exact methods [31–33]. • Storage capacity and storage costs are different for each customer.
The transportation sector generates about 28% of greenhouse gas • Shortage is not allowed.
emissions, mainly released from fossil fuel burner vehicles [34,35]. In
the traditional supply chain system, Logistics, and Transportation (L&T)
companies mainly focused on minimizing their cost and increasing cus- 2.2. Notation
tomer satisfaction. However, with an increase in global warming con-
cerns, L&T companies incorporated the environmental aspects into their • Sets
operations. One of the ways of incorporating ecological concepts into • 𝐼: Set of customers
the routing decisions is minimizing the fuel consumption considering • 𝐼̄: Set of customers and supplier where 𝐼̄ = 𝐼 ∪ {𝑠0 }
loading weights. In this research, a Green Multi-period inventory rout- • 𝑇 : Set of time periods
ing problem with pickup and split delivery (GMIRPSD) is investigated • 𝑉 : Set of vehicles
where storage capacity and storage costs are different for each of the • Model parameters
customers. So, we can use those customers’ storage capacity that have • 𝑐𝑖,𝑗 : Travel cost associated with edge (𝑖, 𝑗) ∈ 𝐼̄.
more storage capacity or less storage. Therefore, in addition to the de- • 𝑞 𝑣 : Capacity of vehicle 𝑣 ∈ 𝑉
livery of the products, we will also have a pickup from the customers. • 𝑓𝑡𝑣 : Fixed cost of vehicle 𝑣 ∈ 𝑉 in period 𝑡 ∈ 𝑇
These pickups and split deliveries will reduce the storage cost and ulti- • 𝑔𝑖 : Storage capacity of customer 𝑖 ∈ 𝐼
mately reduce the system’s total cost. To the best of our knowledge, this • ℎ𝑝𝑖 : Inventory cost of per unit for customer 𝑖 ∈ 𝐼
is the first attempt to model a green multi-period inventory routing prob- • 𝛾 𝑣 : Carbon dioxide emissions caused by vehicle 𝑣 ∈ 𝑉
lem with pickup and split delivery, including the heterogeneous fleet. To • 𝜂: Cost of loading (pickup and delivery) each bag of flour
conclude, the contributions of this study are as follows: (1) We formulate • Decision variables
a bi-objective inventory routing model with pickup and split delivery • 𝑥𝑣𝑖,𝑗,𝑡 : 1 if the edge (𝑖, 𝑗) ∈ 𝐼̄ is traversed by vehicle 𝑣 ∈ 𝑉 in period
to determine the amount of shipment, vehicles’ route and the amount 𝑡 ∈ 𝑇 , and 0 otherwise;
of pickup and delivery at each node; (2) we evaluate the performance • 𝑦𝑣𝑖,𝑗,𝑡 : Amount of product carried by vehicle 𝑣 ∈ 𝑉 on edge (𝑖, 𝑗) ∈
of the proposed model with and without the pickup and split delivery 𝐼̄.
options to better represent the efficacy of the model; (3) we conduct a • 𝑏𝑣𝑖,𝑡 : Amount of delivery to customer 𝑖 ∈ 𝐼 by vehicle 𝑣 ∈ 𝑉 in
case study using real data from a flour supply chain company in Tehran, period 𝑡 ∈ 𝑇
Iran; (4) We provide post-analysis insights which help decision-makers • 𝑎𝑣𝑖,𝑡 : Amount of pickup from customer 𝑖 ∈ 𝐼 by vehicle 𝑣 ∈ 𝑉 in
to decrease the amount of CO2 emission with a small sacrifice in the period 𝑡 ∈ 𝑇
supply chain cost. This can help companies to move towards a sustain- • 𝑟𝑖,𝑡 : Amount of on-hand inventory of customer 𝑖 ∈ 𝐼 in period
able supply chain. The remainder of this paper is organized as follows: 𝑡∈𝑇
Section 2 provides problem definition, notations, and mathematical for- • 𝑢𝑣𝑖,𝑡 : Auxiliary variable associated with node 𝑖 ∈ 𝐼̄ in period 𝑡 ∈ 𝑇
mulation of the problem. Section 3 first demonstrates the benefit of us- for removing the sub-tour of vehicle 𝑣 ∈ 𝑉

65
M. Mahjoob, S.S. Fazeli, L.S. Tavassoli et al. Sustainable Operations and Computers 2 (2021) 64–70

2.3. Mathematical formulation Table 1

Vehicle’s characteristics.

Vehicle 𝑓𝑣 𝑞𝑣 𝛾𝑣
(
∑ ∑∑ ∑ ∑
Min 𝑍1 = 𝑓𝑡𝑣 𝑥𝑣𝑠 + 𝑐𝑖,𝑗 𝑥𝑣𝑖,𝑗,𝑡 + 1 12 240 12
0 ,𝑖,𝑡 2 14 300 14
𝑡∈𝑇 𝑖∈𝐼 𝑣∈𝑉 (𝑖,𝑗)∈𝐼̄ 𝑣∈𝑉
)
∑∑ ∑
𝜂(𝑎𝑣𝑖,𝑡 + 𝑏𝑣𝑖,𝑡 ) + ℎ𝑖 𝑟𝑖,𝑡 (1) Table 2
𝑖∈𝐼 𝑣∈𝑉 𝑖∈𝐼 Storage cost of each customer.

∑ ∑ ∑ Customer 1 2 3 4 5
𝑍2 = 𝛾 𝑣 𝑐𝑖,𝑗 𝑥𝑣𝑖,𝑗,𝑡 (2)
Storage 0.25 0.65 0.5 0.1 0.53
𝑡∈𝑇 (𝑖,𝑗)∈𝐼̄ 𝑣∈𝑉
Capacity 200 200 100 400 200

s.t. Table 3
∑ Demand at each period.
𝑥𝑣𝑖,𝑗,𝑡 ≤1 ∀𝑖 ∈ 𝐼̄, 𝑡 ∈ 𝑇 , 𝑣 ∈ 𝑉 , (3)
𝑗∈𝐼̄ Periods
Parameters
1 2 3
∑ ∑
𝑥𝑣𝑖,𝑗,𝑡 − 𝑥𝑣𝑘,𝑖,𝑡 = 0 ∀𝑖 ∈ 𝐼̄, 𝑡 ∈ 𝑇 , 𝑣 ∈ 𝑉 , (4) 1 63 44 38
𝑗∈𝐼̄ 𝑘∈𝐼̄ 2 62 68 42
3 56 65 49
4 45 52 39
𝑦𝑣𝑖,𝑗,𝑡 ≤ 𝑞 𝑣 𝑥𝑣𝑖,𝑗,𝑡 ∀(𝑖, 𝑗) ∈ 𝐼̄, 𝑡 ∈ 𝑇 , 𝑣 ∈ 𝑉 , (5) 5 63 58 63

Table 4
∑ ∑
𝑦𝑣𝑗,𝑖,𝑡 − 𝑦𝑣𝑖,𝑘,𝑡 = 𝑏𝑣𝑖,𝑡 − 𝑎𝑣𝑖,𝑡 ∀𝑖 ∈ 𝐼̄, 𝑡 ∈ 𝑇 , 𝑣 ∈ 𝑉 (6) Distance between nodes.
𝑗∈𝐼̄ 𝑘∈𝐼̄
Customer 0 1 2 3 4 5

( ) 0 0 45 72 90 92 85
∑ ∑ ∑
𝑟𝑖,𝑡−1 − 𝑟𝑖,𝑡 + 𝑦𝑣𝑗,𝑖,𝑡 − 𝑦𝑣𝑖,𝑘,𝑡 = 𝑑𝑖,𝑡 1 45 0 40 85 92 92
2 72 40 0 50 67 73
𝑣∈𝑉 𝑗∈𝐼̄ 𝑘∈𝐼̄
3 90 81 50 0 20 32
∀𝑖 ∈ 𝐼̄, 𝑡 ∈ 𝑇 (7) 4 92 92 67 20 0 14
5 85 92 73 32 14 0

𝑢𝑣𝑖,𝑡 − 𝑢𝑣𝑗,𝑡 + |𝐼 |𝑥𝑣𝑖,𝑗 ≤ |𝐼 | − 1 ∀(𝑖, 𝑗) ∈ 𝐼, 𝑡 ∈ 𝑇 , 𝑣 ∈ 𝑉 , (8)

3.1. Explanatory example

To evaluate the performance of the proposed model, we consider a

𝑟𝑖,𝑡 ≤ 𝑔𝑖 ∀𝑖 ∈ 𝐼, 𝑡 ∈ 𝑇 , (9) supplier that is committed to meet five different customers’ demands
within a given period (e.g., three months). We assume that there are
two vehicles available at the supplier depot with limited capacity. De-
𝑥𝑣𝑖,𝑗,𝑡 ∈ {0, 1} ∀(𝑖, 𝑗) ∈ 𝐼̄, 𝑡 ∈ 𝑇 , 𝑣 ∈ 𝑉 , (10) tails about the capacity and cost rates, as well as the amount of carbon
dioxide emissions for each type of available vehicle, are given in Table 1.
The initial inventory of all the customers is considered as zero.
𝑟𝑖,𝑡 , 𝑦𝑣𝑖,𝑘,𝑡 ≥ 0 ∀𝑖 ∈ 𝐼̄, 𝑡 ∈ 𝑇 , 𝑣 ∈ 𝑉 (11) The cost of loading of each unit of product in each of the customers
node is assumed to be 𝜂 = 0.05. The cost of storing each unit of product
Eq. (1) shows the first objective function consisting of the vehicles’ in for each customer is given in Table 2.
fixed cost, transportation cost, inventory cost, and the cost of loading the Table 3 provides the travel distances between the nodes in the distri-
product (pickup and delivery). The second objective function indicates bution network. The data regarding the demand for each of the customer
the amount of carbon dioxide emissions during the planning horizon. is given in Table 4.
Constraints (3) guarantee that each customer is visited at most once To demonstrate the performance of the proposed model, we solve
by a vehicle in a given period. Constraints (4) guarantee that the in- the sample problem in two situations: ‘A’) when pickup is not allowed
degree of each node is equal to its out-degree. Constraints (5) ensure that (𝑎𝑖 = 0); and ‘B’) when pickup is allowed (𝑎𝑖 ≠ 0). Situation ‘B’ is what
when there is no route between two nodes, the amount of the product we present as the mathematical formulation in Section (2). For situation
shipped between these two nodes is zero, and also, the shipped amount ‘A’, where pickup is not allowed, after removing the pickup term from
is less than or equal to the capacity of the vehicle. Constraints (6) and the constraints (6) we have:
(7) balance inventory levels for any combination of customers and peri- ∑ ∑
𝑦𝑣𝑗,𝑖,𝑡 − 𝑦𝑣𝑖,𝑘,𝑡 = 𝑏𝑣𝑖,𝑡 (12)
ods. Constraints (8) remove the sub-tours in any feasible solution. Con-
𝑗∈𝐼̄ 𝑘∈𝐼̄
straints (9) guarantee that the on-hand balance for each customer and in
each period is less its storage capacity. Constraints (10) and (11) define and since the delivery amount (𝑏𝑣𝑖,𝑡 ) for each 𝑖 ∈ 𝐼, 𝑣 ∈ 𝑉 , 𝑡 ∈ 𝑇 is always
variable restriction. a non-negative value, Eq. (12) is converted as follows: constraints (6) we
have:
∑ ∑
3. Model performance and methodology 𝑦𝑣𝑗,𝑖,𝑡 − 𝑦𝑣𝑖,𝑘,𝑡 ≥ 𝑏𝑣𝑖,𝑡 (13)
𝑗∈𝐼̄ 𝑘∈𝐼̄

In this section, we first demonstrate the performance of the proposed Constraints (13) which indicate the amount of flour delivered to cus-
model using an explanatory example. tomer 𝑖 in period 𝑡, ensures that the feasible solutions do not contain

66
M. Mahjoob, S.S. Fazeli, L.S. Tavassoli et al. Sustainable Operations and Computers 2 (2021) 64–70

Fig. 1. The generated routes the lowest inventory-routing in each period by solving the model for the situation ‘A’ subject to the first objective.

Fig. 2. The generated routes with the lowest CO2 in each period based on solving the model for the situation ‘A’ subject to the second objective.

Fig. 3. The generated routes with the lowest inventory-routing cost in each period based on solving the model for situation ‘B’ subject to the first objective function.

sub-tour. In this case, constraints (8) will be a redundant constraint. Table 5

Therefore, when the pickup is not allowed, constraints (6) is converted Comparison of objective values and their components.
to constraints (13), and constraints (8) is eliminated, meaning that situa- A B
tion A includes all the constraints in Section 2, plus constraints (13) and Parameters
𝑍1 𝑍2 𝑍1 𝑍2
excluding constraints (6) and (8). The results are shown in Figs. 1–4. In
these figures, the node 0 denotes the distributor, and the nodes 1 to 5 𝑍1 836.04 914.51 824.54 904.34
indicate each of the customers. We used two types of arrows to show 𝑍2 860.4 838.2 863.4 822.9
Fixed cost 40 40 40 40
the routes traveled by each of the vehicles; the solid arrow is related to Variable cost 605 607 607 611
vehicle 𝑣1 , and the dashed arrow represents vehicle 𝑣2 . The amount of Inventory cost 150.69 227.16 132.69 197.49
the shipped products are shown on top of the arrows. Also, the numbers Load/Unload cost 40.35 40.35 44.85 55.85
in green represent the amount of inventory available in the customers’
storage space at the end of each period, and finally, the numbers in
red represent the amount of the pickup from the customers. Figs. 1 and same constraints. Therefore, the model could obtain the lowest amount
3 illustrate the routes generated with the lowest inventory-routing cost of carbon dioxide emissions.
in each period by solving the model for situation ‘A’ and ‘B’ respec- The summary of the results is provided in Table 5.
tively. In this case, as already mentioned, all the constraints in Section The value of the first objective function (𝑍1 ) in the situation ‘B’ is
(2), plus constraints (13) and excluding constraints (6) and (8) are ac- about 1% lower than the situation ‘A’. The difference can be interpreted
tive and must be optimized considering the first objective function. In by looking at the detailed cost in objective functions. In both situations,
Figs. 2 and 4, the second objective function is considered subject to the the vehicles’ fixed cost, loading cost, and distance cost are almost the

67
M. Mahjoob, S.S. Fazeli, L.S. Tavassoli et al. Sustainable Operations and Computers 2 (2021) 64–70

Fig. 4. The generated routes with the lowest inventory-routing cost in each period based on solving the model for situation ‘B’ subject to the first objective function.

same. The difference comes from the lower inventory cost (12%) in the
situation ‘B’ where the pickup is allowed. As it is shown in Fig. 1, at the
end of period 1, customer 4 and 3 have stored 91 and 45 unit of products.
However, in Fig. 3, where the pickup is allowed, because the inventory
cost of customer 4 is 80% less than the customer 3, 136 products are
stored in customer 4 at the end of period 1. Although this amount is
more than the demand for customer 4, the effect of this policy could be
observed in the next period. As it is illustrated in Fig. 3, the vehicle 2
leaves customer 3 carrying 76 products to visit customer 4. In customer
4, 45 units of products are loaded into vehicle 2, and no delivery hap-
pened. A total of 121 units of products are delivered to the customer 5
by vehicle 2. The reason for this transition is the lower inventory cost
(45 × 0.1 = 4.5) in customer 5, comparing to customer 4 (45 × 0.5 = 22.5).
From the ecological perspective, we observe lower CO2 emissions in the
situation ‘B’ compared to Figs. 1–5 ‘A’. As it is shown in Fig. 2 in period
1, the route for vehicle 1 and 2 are 0-5-4-3-2-0 and 0-2-1-0, respectively.
However, in Fig. 4, the routes for 0-5-4-3-1-2-0 and 0-1-0, which leads
to 37 shorter distance units and, consequently, 12 units less CO2 emis-
sions. In conclusion, in this example, we demonstrated two important
features of the proposed model (1) Pickup (2) Split Delivery. These two
features together could decrease the total cost of the system as well as
Fig. 5. Emission level vs. total cost respect to different values for Lp-metric
CO2 emissions.
coefficient.

4. Experiments and results

To showcase the application of the proposed model, we collected

1. 0 ≤ 𝑍 ∗ ≤ 1
data related to a flour supply chain in Tehran, Iran. Fig. 5 shows the ge-
2. 𝑍1min ≤ 𝑍1 ≤ 𝑍1max
ographic distribution of the bakeries. Since the model is bi-objective, it is
necessary to choose a proper approach to solve this problem efficiently. 3. 𝑍2min ≤ 𝑍2 ≤ 𝑍2max
There are three methods in the literature to deal with multi-objective
problems. (1) Priori method (Lp-metric), where decision-makers re-
veal their priorities before settling the model. In this case, the multi- where 𝑍 ∗ denotes the optimal solution of the Lp-metric objective. By
objective problem is converted to a single objective [36–38]. (2) Poste- changing the Lp-metric coefficient, we provide a set of solutions for
rior method, which is based on solving different objectives at the same decision-makers to choose a proper solution based on their preference.
time to obtain a Pareto set of solutions [39,40]. (3) Interactive approach, Fig. 6 shows the trade-off between the CO2 emissions and the total cost
where the decision-makers continuously search for the preferable solu- of the system. As we increase w, we observe the increase in emissions
tion [5,41,42]. and vise versa. Finally, we considered a balanced weight (𝑤 = 0.5) for
For the proposed model in this work, we choose the Priori method. objective functions and solved 12 instances with different characteris-
Considering 𝑍1 and 𝑍2 as the first and second objective functions in our tics. In Table 6, the first four columns indicate the instance number, the
problem, based on the Lp-metric method, the model should be optimized number of bakeries, vehicles, and periods respectively. The next col-
for each objective separately. We define 𝑍1∗ and 𝑍2∗ as the optimal value umn represents the pay-off table, where at each run, one objective is
for the objective functions and w as the Lp-metric coefficient such that selected as the primary objective, and the value of other objective is
0 ≤ 𝑤 ≤ 1. The Lp-metric objective function is defined as follows: calculated after the optimal solution is obtained. Column 7 reports the
( ) results based on the Lp-metric approach. Columns 6 and 8 indicate the
( 𝑍1 − 𝑍1∗ ) ( 𝑍2 − 𝑍2∗ )
Min𝑍 = 𝑤 ⋅ + (1 − 𝑤) ⋅ run-time in terms of seconds for the pay-off and Lp-metric approaches,
𝑍1 − 𝑍1
max ∗ 𝑍2 − 𝑍2
max ∗
respectively. All the experiments were performed on a computer with
Where 𝑍1max and 𝑍1max are the objective value of the first and second Intel(R) Core(TM) i7-9750H CPU@ 2.6 GHz with 32 GB of RAM. All the
objective functions with respect to the 𝑍2∗ and 𝑍1∗ . So, by using the Lp- experiments were solved using Cplex 11.0 software. We limit the run-
metric objective, we have a mixed-integer linear programming where time to two hours for all instances. As expected, with the increase in the
the results satisfy the following conditions: size of the problem, we see an increase in the run-time.

68
M. Mahjoob, S.S. Fazeli, L.S. Tavassoli et al. Sustainable Operations and Computers 2 (2021) 64–70

Table 6
Computational results for the generated instances.

by a fleet of vehicles with different capacities in a given time horizon.

The pickups and split deliveries enable the distribution system to reduce
the flour storage cost and, consequently, the total cost of the system.
The proposed model also considers the environmental aspects to reduce
greenhouse gas emissions from transportation operations. We evaluated
the performance of the model using a case study considering a flour
distribution supply chain in a part of Tehran, Iran. The results showed
that by enabling the pickup and split delivery, we could significantly de-
crease the cost of the flour supply chain. As a solution methodology, we
implemented a Priori method to provide a set of solutions for decision-
makers to select the desired solutions based on their preferences. Be-
sides, we ran the experiments for large-scale instances to quantify the
performance of the proposed model under different scenarios. A poten-
tial research extension could be to design a novel heuristic algorithm to
solve large-scale instances efficiently. Another future research could be
involving the demand uncertainty in the proposed model. In addition,
the shortage in distribution centers could be allowed to capture more
realistic cases.

Declaration of Competing Interest

Fig. 6. The location of the bakeries in a part of Tehran, Iran shown with red
markers.
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence
the work reported in this paper.
5. Conclusion
References
In this research, a new green multi-period inventory routing prob-
[1] M. Mahmoudi, I. Parviziomran, Reusable packaging in supply chains: a review of
lem with pickup and split delivery for distributing flour among bakeries environmental and economic impacts, logistics system designs, and operations man-
is proposed. Flour bags are shipped from the flour mill to the bakeries agement, Int. J. Prod. Econ. (2020) 107730.

69
M. Mahjoob, S.S. Fazeli, L.S. Tavassoli et al. Sustainable Operations and Computers 2 (2021) 64–70

[2] M. Boroun, S. Ramezani, N. Vasheghani Farahani, E. Hassannayebi, S. Abolmaali, [23] G. Berbeglia, J.-F. Cordeau, G. Laporte, Dynamic pickup and delivery problems, Eur.
M. Shakibayifar, An efficient heuristic method for joint optimization of train schedul- J. Oper. Res. 202 (1) (2010) 8–15.
ing and stop planning on double-track railway systems, INFOR Inf. Syst. Oper. Res. [24] A. Tafreshian, N. Masoud, Trip-based graph partitioning in dynamic ridesharing,
58 (4) (2020) 652–679. Transp. Res. Part C Emerg. Technol. 114 (2020) 532–553.
[3] S. Lashkari, Y. Wu, M.E. Petering, Sequencing dual-spreader crane operations: math- [25] G. Iassinovskaia, S. Limbourg, F. Riane, The inventory-routing problem of return-
ematical formulation and heuristic algorithm, Eur. J. Oper. Res. 262 (2) (2017) able transport items with time windows and simultaneous pickup and delivery in
521–534. closed-loop supply chains, Int. J. Prod. Econ. 183 (2017) 570–582.
[4] J.-F. Cordeau, D. Laganà, R. Musmanno, F. Vocaturo, A decomposition-based heuris- [26] R.G. Van Anholt, L.C. Coelho, G. Laporte, I.F. Vis, An inventory-routing problem with
tic for the multiple-product inventory-routing problem, Comput. Oper. Res. 55 pickups and deliveries arising in the replenishment of automated teller machines,
(2015) 153–166. Transp. Sci. 50 (3) (2016) 1077–1091.
[5] E. Taghizadeh, M. Abedzadeh, M. Setak, A multi objective reliable location- [27] T.J. Ai, V. Kachitvichyanukul, A particle swarm optimization for the vehicle routing
inventory capacitated disruption facility problem with penalty cost solve with ef- problem with simultaneous pickup and delivery, Comput. Oper. Res. 36 (5) (2009)
ficient meta historic algorithms, arXiv preprint arXiv:1711.09400(2017). 1693–1702.
[6] L. Qin, L. Miao, Q. Ruan, Y. Zhang, A local search method for periodic inventory [28] A. Subramanian, L.M.d.A. Drummond, C. Bentes, L.S. Ochi, R. Farias, A parallel
routing problem, Expert Syst. Appl. 41 (2) (2014) 765–778. heuristic for the vehicle routing problem with simultaneous pickup and delivery,
[7] M. Torabbeigi, G.J. Lim, S.J. Kim, Drone delivery scheduling optimization consid- Comput. Oper. Res. 37 (11) (2010) 1899–1911.
ering payload-induced battery consumption rates, Journal of Intelligent & Robotic [29] N. Bianchessi, G. Righini, Heuristic algorithms for the vehicle routing problem with
Systems 97 (3) (2020) 471–487. simultaneous pick-up and delivery, Comput. Oper. Res. 34 (2) (2007) 578–594.
[8] A. Fallah-tafti, M.A. Vahdatzad, A. Sadeghieh, A comprehensive mathematical model [30] F.A.T. Montané, R.D. Galvao, A tabu search algorithm for the vehicle routing prob-
for a location-routing-inventory problem under uncertain demand: a numerical il- lem with simultaneous pick-up and delivery service, Comput. Oper. Res. 33 (3)
lustration in cash-in-transit sector, Int. J. Eng. 32 (11) (2019) 1634–1642. (2006) 595–619.
[9] A. Federgruen, D. Simchi-Levi, Analysis of vehicle routing and inventory-routing [31] J. Tang, Y. Yu, J. Li, An exact algorithm for the multi-trip vehicle routing and
problems, in: Handbooks in Operations Research and Management Science, 8, 1995, scheduling problem of pickup and delivery of customers to the airport, Transp. Res.
pp. 297–373. Part E Logist. Transp. Rev. 73 (2015) 114–132.
[10] P. Jaillet, J.F. Bard, L. Huang, M. Dror, Delivery cost approximations for inven- [32] M. Casazza, A. Ceselli, R.W. Calvo, A route decomposition approach for the single
tory routing problems in a rolling horizon framework, Transp. Sci. 36 (3) (2002) commodity split pickup and split delivery vehicle routing problem, Eur. J. Oper. Res.
292–300. 289 (3) (2021) 897–911.
[11] J.F. Bard, L. Huang, P. Jaillet, M. Dror, A decomposition approach to the inventory [33] G. Dong, J. Tang, K.K. Lai, Y. Kong, An exact algorithm for vehicle routing and
routing problem with satellite facilities, Transp. Sci. 32 (2) (1998) 189–203. scheduling problem of free pickup and delivery service in flight ticket sales compa-
[12] H. Ebrahimi, M. Tadic, Optimization of dangerous goods transport in urban zone, nies based on set-partitioning model, J. Intell. Manuf. 22 (5) (2011) 789–799.
Decis. Mak. Appl. Manag. Eng. 1 (2) (2018) 131–152. [34] S.S. Fazeli, S. Venkatachalam, R.B. Chinnam, A. Murat, Two-stage stochastic choice
[13] P.S. Barma, J. Dutta, A. Mukherjee, A 2-opt guided discrete antlion optimization modeling approach for electric vehicle charging station network design in urban
algorithm for multi-depot vehicle routing problem, Decis. Mak. Appl. Manag. Eng. communities, IEEE Transactions on Intelligent Transportation Systems (2020) 1–16,
2 (2) (2019) 112–125. doi:10.1109/TITS.2020.2979363.
[14] A.M. Sarmiento, R. Nagi, A review of integrated analysis of production-distribution [35] S.S. Fazeli, S. Venkatachalam, J.M. Smereka, Efficient algorithms for au-
systems, IIE Trans. 31 (11) (1999) 1061–1074. tonomous electric vehicles’ min-max routing problem, arXiv preprint
[15] H. Andersson, A. Hoff, M. Christiansen, G. Hasle, A. Løkketangen, Industrial aspects arXiv:2008.03333(2020b).
and literature survey: combined inventory management and routing, Comput. Oper. [36] S.M. Mirzapour Al-e hashem, Y. Rekik, E.M. Hoseinhajlou, A hybrid l-shaped method
Res. 37 (9) (2010) 1515–1536. to solve a bi-objective stochastic transshipment-enabled inventory routing problem,
[16] W.J. Bell, L.M. Dalberto, M.L. Fisher, A.J. Greenfield, R. Jaikumar, P. Kedia, Int. J. Prod. Econ. 209 (2019) 381–398.
R.G. Mack, P.J. Prutzman, Improving the distribution of industrial gases with an [37] P. Abbasian, N. Mahdavi-Amiri, H. Fazllolahtabar, Modeling and optimizing of a
on-line computerized routing and scheduling optimizer, Interfaces 13 (6) (1983) non-constrained multi-objective problem having multiple utility functions using the
4–23. Bayesian theory, J. Inf. Sci. Comput. Technol. 4 (3) (2015) 332–342.
[17] J. Cho, G.J. Lim, S.J. Kim, T. Biobaku, Liquefied natural gas inventory routing prob- [38] P. Abbasian, N. Mahdavi-Amiri, H. Fazlollahtabar, Multiple utility constrained mul-
lem under uncertain weather conditions, Int. J. Prod. Econ. 204 (2018) 18–29. ti-objective programs using Bayesian theory, J. Ind. Eng. Int. 14 (1) (2018) 111–118.
[18] F. Al-Khayyal, S.-J. Hwang, Inventory constrained maritime routing and scheduling [39] M. Tavana, K. Khalili-Damghani, D. Di Caprio, Z. Oveisi, An evolutionary computa-
for multi-commodity liquid bulk, part I: applications and model, Eur. J. Oper. Res. tion approach to solving repairable multi-state multi-objective redundancy alloca-
176 (1) (2007) 106–130. tion problems, Neural Comput. Appl. 30 (1) (2018) 127–139.
[19] A. De, S.K. Kumar, A. Gunasekaran, M.K. Tiwari, Sustainable maritime inventory [40] L.S. Tavassoli, N. Sakhavand, S.S. Fazeli, Integrated preventive maintenance
routing problem with time window constraints, Eng. Appl. Artif. Intell. 61 (2017) scheduling model with redundancy for cutting tools on a single machine, Eng. Tech-
77–95. nol. Appl. Sci. Res. 10 (6) (2020) 6542–6548.
[20] J. Alegre, M. Laguna, J. Pacheco, Optimizing the periodic pick-up of raw materials [41] S. Rajendran, A. Ansaripour, M. Kris Srinivasan, M.J. Chandra, Stochastic goal pro-
for a manufacturer of auto parts, Eur. J. Oper. Res. 179 (3) (2007) 736–746. gramming approach to determine the side effects to be labeled on pharmaceutical
[21] V. Gaur, M.L. Fisher, A periodic inventory routing problem at a supermarket chain, drugs, IISE Trans. Healthc. Syst. Eng. 9 (1) (2019) 83–94.
Oper. Res. 52 (6) (2004) 813–822. [42] A. Molavi, J. Shi, Y. Wu, G.J. Lim, Enabling smart ports through the integration of
[22] K.P. Mubiru, K.B. Buhwezi, O.P. Lating, The joint location inventory replenishment microgrids: a two-stage stochastic programming approach, Appl. Energy 258 (2020)
problem at a supermarket chain under stochastic demand, J. Ind. Eng. Manag. Sci. 114022.
2019 (1) (2019) 161–178.

70