See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/282242551

A metaheuristic ACO to solve the multi-depot vehicle routing problem with

backhauls

Article  in  International Journal of Industrial Engineering and Management · January 2015

CITATIONS READS

4 193

4 authors:

Jhon jairo Santa Mauricio Granada-Echeverri

Universidad Tecnológica de Pereira Universidad Tecnológica de Pereira
8 PUBLICATIONS   37 CITATIONS    109 PUBLICATIONS   395 CITATIONS   

SEE PROFILE SEE PROFILE

John Willmer Escobar César Augusto Peñuela

Pontificia Universidad Javeriana - Cali Universidad Libre Seccional Pereira
94 PUBLICATIONS   576 CITATIONS    12 PUBLICATIONS   66 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Algorithms for Multi Depot Electric Vehicle Routing Problems View project

Electric Vehicle Routing View project

All content following this page was uploaded by John Willmer Escobar on 18 March 2016.

The user has requested enhancement of the downloaded file.

 International Journal of Industrial Engineering and Management (IJIEM), Vol. 6 No 2, 2015, pp. 49-58
Available online at www.iim.ftn.uns.ac.rs/ijiem_journal.php
ISSN 2217-2661

UDK: 656.07

A Metaheuristic ACO to Solve the Multi-Depot Vehicle

Routing Problem with Backhauls
Jhon Jairo Santa Chávez
Universidad Libre Seccional Pereira, Pereira (Colombia), [email protected], Universidad
Tecnológica de Pereira, Pereira (Colombia),[email protected]
Mauricio Granada Echeverri
Universidad Tecnológica de Pereira, Pereira (Colombia),[email protected]
John Willmer Escobar*
Pontificia Universidad Javeriana Cali, Cali (Colombia),[email protected]
César Augusto Peñuela Meneses
Universidad Libre Seccional Pereira, Pereira (Colombia), [email protected]

Received (11.05.2015.); Revised (17.06.2015.); Accepted (03.07.2015.)

Abstract
In this paper, a metaheuristic algorithm based on the ant colony optimization is presented to solve the
multi-depot vehicle routing problem with delivery and collection of package. Each performed route
consists of one sub-route in which only the delivery task is done, in addition to one sub-route in which
only the collection process is performed. The proposed algorithm tries to find the best order to visit the
customers at each performed route. In addition, the proposed approach determines the best
connection between the sub-routes of delivery and collection, in order to obtain a global solution with
the minimum travelling cost. The efficiency of the proposed algorithm has been evaluated by
considering a set of instances adapted from the literature. The computational results have been
compared with a greedy heuristic algorithm based on the nearest neighborhood approach. Finally,
conclusions and suggestions for future works are presented.
Key words: Ant Colony, Backhauls, Combinatorial Optimization,Computational Simulation, Multi-
Depot Vehicle Routing Problem.

1. INTRODUCTION consequence of merging them. In this case, the routes

are merged to create a route that
The transportation and logistics companies are facing will be attended by only one vehicle. In all possible
demanding situations with fewer resources. Currently, scenarios, the capacity of the vehicle must be ensured.
an important aspect is to optimize the use of existing The Multiple-Depot Vehicle Routing Problem with
resources in order to perform specific tasks. Many of Backhauls (MDVRPB) could be defined as follows:
these tasks in real-world problems can be solved by a Given two sets of predefined customers, where is the
network representation containing nodes, connections, set of Linehaul customers (customers to be delivered
variables, parameters, and their interactions. In recent products), and the set of Backhaul customers
decades, there have been various proposals for solving (customers to be collected products). In addition, given
vehicle routing problems with backhauls by considering a set of geographically dispersed depots; the objective
only capacity constraints. Toth and Vigo [1] propose the is to consider the routes to be performed from the
first approach to the formulation of a vehicle routing selected depots to the customers by a fleet of
problem based on a fuel distribution problem. In Clarke homogeneous vehicles in order to satisfy the demand of
and Wright [2], the first effective heuristic algorithm for the customers (products to be collected or products to
solving the Capacity Vehicle Routing Problem (CVRP) be delivered). In such case, the vehicles must attend
is proposed. first the customers with delivery requirements before
This method starts with a set of vehicles located at the the customers with collection requirements.
depot. Some of these vehicles are selected to deliver The features of the customers, depots and vehicles, as
the cargo to the customers through predefined routes. well as the different operating constraints on the
Then, two routes are selected iteratively in order to performed routes, leads to different variants of the
determine if a lower distance can be generated as a problem: The Vehicle routing Problem with Backhaul
(VRPB) in which only one depot is considered, the

IJIEM
50 Santa Chávez et al.

Mixed Vehicle routing Problem with Backhaul (MVRPB) depots is considered. Two algorithms based on ant
in which a fleet of homogeneous capacity vehicles is colony for the multi depot vehicle routing problem with
concerned, and the VRPB with time windows multiple depots and mixed pickup and delivery are
(VRPBTW) in which the expected time to serve the presented in [7]-[8]. Finally, genetic algorithms for
clients is constrained. The MDVRPB is considered as solving the MDVRPB have been proposed by Chunyu &
NP-hard, and has not been deeply explored in the Xiaobo [9] and Chunyu [10].
literature. The variants of the MDVRPB can be This paper proposes an algorithm based on an Ant
classified into three categories: (i) simultaneous Colony System to solve the MDVRPB with collection of
collecting and dispatching of products; (ii) collecting products exclusively after delivery of products. The
first, following the delivery of products; and (iii) efficiency of the proposed algorithm has been
collecting after of the delivering process. compared in 33 instances taken from the literature,
Recently, several algorithms have been proposed to proposed by Salhi and Nagy [4].
solve the VRPB. In [1], an exact model for the These instances have been used to test the two first
formulation of the VRPB in a fuel distribution scenario is categories of the MDVRPB (simultaneous collecting
proposed. The first heuristic algorithm for VRPB has and dispatching of products and collecting first,
been proposed by Deif and Bodin [3]. This algorithm is following the delivery of products). However, there is no
an extension of the known method of saving method relevant literature on the subject of the variant solution
proposed by Clarke and Wright [2]. The algorithm starts of collecting products after of the delivering process.
with a feasible solution, where each customer is visited Therefore, for comparison purposes, we have
by a different route. implemented a greedy algorithm based on the nearest
Then, the routes are combined iteratively by neighbor idea. The paper is organized as follows:
considering the possible reduction in the objective Section 2 provides a brief description of the MDVRRPB.
function, i.e. the traveling cost of the route that can be Section 3 shows described the metaheuristic Ant
achieved by serving different type of customers with the Colony System. In Section 4, the proposed algorithm is
same vehicle. The results show that the evolution of the introduced. Finally, in Sections 5 and 6, we present the
algorithm is affected by the constraint of precedence of analysis of computational results and the conclusions
customers with delivery requirements. This fact for future research, respectively.
substantially reduces the number of feasible links
between different types of customers. Deif and Bodin 2. PROBLEM STATEMENT
[3] experimentally show that the best results are
obtained when mixed routes are initially constructed. The vehicle routing problem with Backhaul, is an
Therefore, [3] propose a modification of definition of extension of the well-known VRP involving demand
savings in order to penalize the edges connecting nodes and collection of products [1]. The VRPB can be
customers of different types. defined as the problem of determining a set of routes to
Thus, avoiding the union of backhauling customers at visit every customer concerning the following
the end of the algorithm. The savings generates with constrains: each vehicle develops a single route; each
route starts and ends at the depot; the routes must
the use of the edge 'is defined in equation (1). firstly deliver products to customers defined as
Linehaul, and finally pick up products from customers
(1) defined as Backhaul. The MDVRPB is an extension of
VRPB by considering potential depots from which
we can generate different routes. In the MDVRPB all
Where is an estimate of the maximum savings deposits are not necessarily used. The MDVRPB can
between a Linehaul customer ( ) and a Backhaul be formulated as a Graph Theory Problem. Let
customer ( ), and is a penalty factor set to be a complete undirected graph, where
. Deif and Bodin in [3] have tested the proposed is the set of vertices, and is the set of
algorithm on cases randomly generated with instances arcs. The set is partitioned into two subsets: the set
between 100 and 300 customers, and a percentage of of customers , and the set of potential
backhaul customers among 10% to 50% of the total of depots . Additionally, the set is
customers. The Multi Depot Vehicle Routing Problem divided into a subset of Linehaul nodes (Linehaul
with Backhauls (MDVRPB) is an extension of the well- customers - ), and the Backhaul nodes (Backhaul
known VRPB by considering several depots, which are customers – ). Therefore, .
not necessarily used. Salhi and Nagy [4] propose a
Each customer has a nonnegative amount of
heuristic approach to solve the MDVRPB. The algorithm
is based on the idea of "Border Customers", defined as product to be delivered or to be picked
the customers geographically located in a middle point up . Each depot has a fictitious demand, i.e.
between two depots. Min [5] introduced the MDVRPB , with . A set of identical vehicles with
by considering collection process after delivery of a given capacity is initially placed at each depot. It
products. must be clarified that all vehicles are not necessarily
A unified heuristic for different vehicle routing problems used. For each edge merging the vertices
with backhaul is presented in [6]. In this work, the
version of mixed pickup and delivery with multiple with the vertices , a nonnegative cost is associated,

IJIEM
Santa Chávez et al. 51

where for each arc . As the constrained by equations (5) and (6) for linehaul and
symmetrical condition is assumed, for each arc backhaul customers, respectively. Finally, linehaul
customers must be addressed before the backhaul
we assume that = , where . The
customers (7).
main goal of the MDVRPB is to find the minimum cost
Solving the mathematical model (1) – (7) by exact
(traveled distance) of routes, where the following methods, the computational time grows exponentially
conditions must be satisfied: as the number of vertices V is increased. For this
reason, metaheuristic algorithms based on Ant Colony
a) Each vehicle must start from a depot , and System appear as an alternative for the solution of
return to the same depot ; MDVRPB.
b) Each customer must be visited exactly once.
c) The sum of the demands of customers 3. METAHEURÍSTIC ALGORITHM BASED ON
belonging to a determined route must not ANT COLONY SYSTEM
exceed the vehicle capacity ; The Ant Colony metaheuristic is based on the natural
d) For each performed route, the Linehaul behavior of ants searching food. The logical tendency of
customers must precede the Backhaul each ant is to reduce the effort and time required to
customers in the sequence; gather food. This goal is achieved by reducing the
e) The flow between depots is not allowed. distance between two specific points for collecting food.
f) The objective is to minimize the total traveled The ants are individuals with relatively simple features.
distance, defined as the sum of the costs of the However, they perform highly complex work in a simple
arcs belonging to the performed routes. [1] way. The success lies in the interaction of many
individuals with the environment and by indirect
A mathematical model for the MDVRPB has been communication between them through chemicals
introduced in [11]. The objective function is calculated substances called pheromones [12].
as the minimization of the total traveled distance by the This behavior is emulated in artificial intelligence to find
performed routes (1). So, the binary variable is set good quality solutions to optimization problems
characterized by a wide space of solutions. The
to 1 when the vehicle passes by the edge .
concept makes the equivalence between the natural ant
Otherwise is set to 0 colony and the artificial system moving within a
computational environment [12].
(1)
3.1 Selecting arcs
The proposed algorithm interprets an ant as the
emulation of a vehicle in the process of performing its
route. An artificial ant begins its travel from a depot
(2) selected randomly to a first customer. The decision of
the next client to be visited is based on heuristic
preferences biased by the distance among nodes and
(3) the emulation of the natural pheromone.
A probabilistic transition rule defines the likelihood that
the ant (one ant for each available vehicle), placed at
(4) node , decides to move to the node (selecting the
arc The rule is defined according to the
equation (8).
(5)

(8)
(6)

(7) Where is the amount of pheromone on the edge

, and is the heuristic information of ,
i.e. the distance to travel from the node i to the node j.
The level of importance on the rule decision from both
the heuristic information and the amount of pheromone
Equations (2) constrain that a single vehicle visits each is given by parameters and , respectively.
customer once. Additionally, the set of equations (3) Generally, these parameters must have optimized at an
impose the flow constraints for each customer. Each
off-line process. Finally, is the set of neighborhood
performed route must be started and finished at the
same depot (4). The capacity of the vehicles is

IJIEM
52 Santa Chávez et al.

nodes of node i, which have not visited yet by the ant routes. The first two versions generally obtain worse
. results than the third variant [13].
Therefore, we consider the process for updating the
3.2 Setting of parameters and pheromone by assigning a fixed amount of pheromone
at each iteration. The equation (11) is generally used to
In particular, if the nearest nodes have more determine this value, where is the number of ants,
probability to be selected (well-known algorithm of the
i.e. the number of alternative of solutions, and is
gradient with multiple starting points). If only the the tour length obtained at the construct solution
level of pheromone is considered, generally obtaining procedure.
solutions with low quality, especially if . Indeed,
this case causes a stagnation of an algorithm, because
all the ants follow the same way generating sub optimal (11)
solutions. A proper update pheromone matrix directly
impacts the diversity in decision-making process by an
4. PROPOSED ALGORITM FOR THE MDVRPB
ant for the following arcs. We have considered that
when an ant travels by an arc a certain level of 4.1 Encoding Solution
pheromone is deposited, which is evaporated with the
time. The problem of MDVRPB is solved by an ant colony
In particular, an additional positive feedback of the algorithm [13]. Then, the implemented encoding is
amount of pheromone is used to reinforce future presented. The distance matrix is constructed from the
components of good solutions. For a better solution, coordinates of the vertices, i.e. the matrix contains the
more pheromone is provided at its arcs. Euclidean distances among depots and customers. In
The equation (9) determines the rule for increasing the order to improve the efficiency in the search process,
the information is arranged in a redundant way. So, the
quantity of pheromone, where is the amount of Figure 1 shows the distance matrix is a three-
pheromone deposited at the arcs visited by the ant k, dimensional array with dimensions
where is the total cost of the solution generated . It is shown that this matrix can
by the ant k, i.e., the total distance of the route be understood as a set of submatrix each one
developed by the vehicle k. The arcs visited by the dedicated to a particular depot. At any submatrix, the
entire ants in the current solution will receive an extra element represents the cost for travel through the
contribution of pheromone. edge merging the customer i with the customer j. In the
plane corresponding to the distances between
customers, for each deposit, the diagonal is greater as
(9) possible, and the no diagonal values are symmetrical.
The Figure 1 sketches the structure of the distance
matrix.
In addition, pheromone evaporation is used to prevent The Visibility Matrix is obtained from each element of
an unlimited increasing of pheromone trails and to the matrix elements of distances. It contains the inverse
forget low quality solutions. The evaporation level is the of the distance. This matrix is not modified during the
same for all pheromone trails, eliminating a predefined execution of the algorithm and it is symmetrical. Again,
percentage of the current value for each arc each plane corresponds to a particular depot, and its
through a predefined rate , with 0 1. diagonal is high enough. The Linehaul and Backhaul
Thus, the evaporation mechanism could exhibit a customers use the same representation because the
behavior more active than the presented in natural algorithm identifies the delivery or pick up of products
environments. The equation (10) describes the general (Figure 2).
rule for updating the pheromone matrix at the edge Finally, we proceed to generate the pheromone matrix
, at the iteration t. with all positions initialized with the same amount of
pheromone. As shown in the Figure 3, all the elements
at the diagonal are zero, and it is initially a symmetric
(10) matrix. Their values are generated with the equation
(11), and the updating process is restricted to thus
elements in the route traveled by each ant.
3.3 Ant System (AS) Note that the proposed algorithm not considers the
distance among depots, since these distances should
In the ant system algorithm, an artificial colony of ants
not be considered in the distance of the performed
cooperate to find good solutions for discrete, static, and
routes.Indeed, the algorithm uses a new vehicle to
dynamic combinatorial optimization problems.
continue visiting other customers. Therefore, the
Three different variants for the AS have been proposed:
visibility matrix or net matrix is replaced by a squared
Ant-Density, Ant-Quantity and Ant-Cycle [13]. In the first
array of random values that provides diversity in the
two variants, the pheromone updating process is
searching of the new depot to be aggregated to the
performed after each selection arc process, while in the
solution. The total number of depots gives the
third variant (Ant-Cycle), the amount of pheromone is
dimension of the visibility matrix, and its values are
only updated once the ants have completed the entire

IJIEM
Santa Chávez et al. 53

initialized using the equation (11). The nodes are A single linehaul route. This route contains
updated only if they belonging to the route traveled by a only linehaul customers. This route is shown in
vehicle. the Figure 7. The performed route is M1-L4-L2-
In addition, a vector of not visited customers is required. L3-L5-M1(Figure 7).
It contains the feasible customers to be selected. A single backhaul route. This type of route is
Initially, it contains Linehaul customers, and each depicted in the Figure 8. It contains only
selected customer is removed from the vector. The first backhaul customers. The performed route is
element of the vector is 1, and the final element is set to M3-B4-B3-B2-B7-M3.
the total number Linehaul customers ( ). Similarly, it
is needed a vector for Backhaul not visited customers. It The alternative of solution is conformed as a
starts with the total Backhaul customers ( ). When a combination of the possible types of routes. In the
feasible customer is selected, it is removed from the proposed algorithm, the construction of a multi-depot
vector. The first element of vector for Backhaul not solution requires that the ant performs as many
visited customers is the total number of Linehaul subroutes as needed to visit all the customers, and
customers + 1, and the final element is set to total each sub-route comprises a tour considering linehaul
number or customers ( ). and backhaul customers. A possible alternative of
The solution vector is also required. Its first element is solution is shown in the Figure 9. The dashed line
set to the number of a depot, which is randomly represents the entire tour performed by the ant, and it is
selected. It is followed by a sequence of customer until equivalent to an alternative of solution.
a new depot has been selected again. This shape is 4.3 Pseudocode for the proposed algorithm
continued until the entire customers have been
selected. However, at each sequence of customers, a Procedure MDVRPB (depots, customers,
Linehaul customer must not appear between two vehicleNumber, vehicleCapacity)
Backhaul customers, and vice versa. Additionally, the
Backhaul customers are visited before the Linehaul a. Calculate distance matrix
customers. The Figure 4 shows an example of a route b. Initialize the Pheromone matrixes
of the proposed solution. The process to construct an c. Initialize two Net matrixes
alternative of solution is repeated until the population of d. Initialize Pheromone contribution array
the algorithm is complete. The number of the ants to e. Initialize Net contribution array
create is a parameter to be established by an off-line f. It = 0
process. While it < 100 do
Finally, an array stores the solution vectors found by ants = 0
different ants. This array represents the population While ants <size(depots) do
matrix, for which every row is able to store only one // Starting MDVRPB procedure
alternative of solution found by an ant. The number of initialDepot = choose(depots)
rows is the number of iterations by the number of If anyLinehaul(customers) then
depots and the number of columns is at least the Process Linehaul customers
number of customers plus the number of depots to If anyBackhaul(customers) then
satisfy the collections and delivery of products. In the Process Backhaul customers
Figure 5, a set of alternative of solutions is shown. At Evaporate(pheromones)
this example, there are three (3) depots, four (4) update(pheromones)
Linehaul customers, and three (3) Backhaul customers. ants++
End While
4.2 Types of constructed routes End While
Let us define the total number of depots as m: i.e.
, the total number of linehaul customers
5. COMPUTATIONAL RESULTS
as nl: L1, L2, ..., Lnl and the number of backhaul 5.1 Instances
customers as nb: B1, B2, ..., Bnb. As a example, the
proposed algorithm performs the following routes for a The proposed algorithm has been evaluated in 33
system with three (3) depots: M1, M2 and M3; with seven instances MDVRPB described by Salhi and Nagy [5].
(7) linehaul customers: L1, L2, L3, L4, L5, L6 and L7; and The respective set has a diverse number of Linehaul
seven (7) backhaul customers: B1, B2, B3, B4, B5, B6 and and Backhaul customers (from 50 to 250), a number of
B7. We can consider the following types of routes: depots is between 2-5, and a homogeneous set of
vehicles is considered. Euclidean distance for each arc
A linehaul – backhaul route. The Figure 6 is performed. The set of benchmark instances has been
describes this type of route. Its main feature is used to solve the problem of allowing MDVRPB
that both the linehaul sequence and the percentages in the delivery and pickup of 10%, 25%
backhaul sequence are reaching the capacity of and 50% (Salhi and Nagy [5] and Ropke and Pisinger
the vehicle. The performed route is M2-L1-L6-L7- [6]). However in the reviewed literature, there is not
B5-B1-B6-M2. found papers that solve the problem MDVRPB with
Backhauls at the end of the routes. Therefore, we have
implemented a simple greedy heuristic based on the

IJIEM
54 Santa Chávez et al.

nearest neighborhood in order to make a comparison decision determines whether the next route will be
with the proposed approach. started from the current depot, or at other depot.
5.2 Description of the Greedy algorithm At each case, the procedure described before is
repeated. The overall process is executed until all
The greedy algorithm is a constructive approach based
customers had been attended.
on the nearest neighbor, i.e. the lower distance to the
current location. Hence, starting from a particular depot,
the route in construction receives the customer with the
lower distance to the depot among the customers not
assigned yet. 5.3 Obtained Results
This rule is sequentially repeated, but the current The proposed algorithm has been implemented in
location regards a last customer added to the solution. MatLab, and computational experiments have been run
So, the feasible customer is evaluated through the on a PC with Core i5 1.4 GHz processor and 8GB of
distance to the last customer in the route in RAM. The Table 1 shows the obtained results in
comparison with the heuristic nearest neighborhood.
consideration. When a backhaul customer is added to
We report the computing time and the values reached
the solution, the next location cannot be a linehaul by the metaheuristic objective proposed. The algorithm
customer, and only backhaul customers are considered, has been executed for 10 runs over 100 iterations,
as well as coming back to the initial depot. A random reporting the average results and the best results of all
runs.

Depot M

Depot 2

Depot 1

Depot M

Depot 2
Depot 1

Figure 1. Encoding of the distance matrix.

Figure 2. Encoding of the Visibility matrix

IJIEM
Santa Chávez et al. 55

Figure 3. Encoding the pheromone matrix

D2 L1 L6 L7 B5 B1 B6 D2 D1 L4 L2 D1 D3 B4 B3 B2 B7 D3
Linehaul-Backhaul route Linehaul route Backhaul route

Figure 4. Encoding the solution Vector

Solution 1 D2 L4 L3 B1 B2 D2 D3 L1 L2 D3 D1 B3 D1
Solution 2 D1 L1 L2 D1 D2 L3 B2 D2 D3 L4 B1 B3 D3

Solution i D3 L1 L2 L3 B1 D3 D1 L4 B2 B3 D1

Solution k D2 L2 L1 D2 D1 L3 B2 D1 D3 L4 B1 B3 D3

Figure 5. Encoding of the Matrix of the solution

B2 L2
B6 B5 L3

Depot 2 L4
L7
L1 L5
Depot 1
L6
Figure 7. Single linehaul route.
Figure 6. linehaul – backhaul route

B7
B7
B7
B7
Depot 3

Figure 8. Single backhaul route.

IJIEM
56 Santa Chávez et al.

The parameters used to execute instances are the [3] Deif I., Bodin L. (1984). “Extension Of The Clarke And Wright
Algorithm For Solving The Vehicle Routing Problem With
following: Number of Iterations: 100, Distance
Backhauling”. in A. E. Kidder, Hrsg., Proceedings of Babson
parameter, = 1, Pheromone parameter, = 3, Conference on Software Uses in Transportation and Logistics
Evaporation effect parameter, = 0.01, contribution Management, pages 75-96, University of Maryland, College
factor of the visited arc = Number of Customers / Best Park, MD 20742, USA, College of Business and Management
Babson Park.
Solution Found.
[4] Salhi S., Nagy G. (1999).“A cluster insertion heuristic for single
and multiple depot vehicle routing problems with backhauling”,
6. CONCLUDING REMARKS Journal of the Operational Research Society, Vol. 50, pp. 1034-
1042.
In this paper, an effective algorithm for the Multi-Depot [5] Min, H., Current, J., Schilling, D., & Current, J, (1992). “The
Vehicle Routing Problem with Backhauls (MDVRPB) is multiple depot vehicle routing problem with backhauling”. Journal
of Business Logistics, Vol. 13, pp. 259-288.
proposed. The development of the proposed algorithm [6] Ropke S., Pisinger D, (2006). “A unified heuristic for a large
has been evaluated by considering benchmarking class of vehicle routing problems with backhauls”. European
instances proposed in the literature. However Journal of Operational Research, Vol. 171, pp. 750–775.
comparison of the results obtained have been [7] Wade, A., Salhi S., (2003). “An ant system algorithm for the
mixed vehicle routing problem with backhauls”, in M.G.C.
performed with a simple heuristic based on the nearest Resende and J.P. de Sousa (eds.): Metaheuristics: Computer
neighborhood due to that there is not algorithms in the Decision-Making, Chapter 33, pp. 699-719, Kluwer.
literature that have considered the same variant of the [8] Wade A., Salhi S. (2001). “An ant system algorithm for the
problem. Computational results show the proposed vehicle routing problem with backhauls”, in MIC’2001 - 4th
Metaheuristic International Conference.
algorithm is capable of producing, within short [9] Chunyu, R., Wang X. (2009),“Study on hybrid genetic algorithm
computational times, several solutions of high quality. In for multi-type vehicles and multi-depot vehicle routing problem
fact, the former algorithm exceeds the results obtained with backhauls”. inSecond International Conference on
by the greedy heuristic based on the nearest Intelligent Computation Technology and Automation, 2009. Vol.
1. IEEE.
neighborhood in an 19.01%. The results suggest that [10] Chunyu, R., Song Z., Wang X. (2009),“Study on Single and
the proposed algorithm could be applied to other routing Mixed Fleet Strategy for Multi-depot Vehicle Routing Problem
problems such as the Multi Depot Vehicle Routing with Backhauls”. In International Conference on Computational
Problem (MDVRP), the Periodic Location Routing Intelligence and Natural Computing, CINC'09. Vol. 1. IEEE.
[11] Rizka R. (2013). “Multiple Depot Vehicle Routing Problem with
Problem (PLRP), the Muti-Depot Vehicle Routing Backhauls Menggunakan Algoritma Clark and Wright dengan 2-
Problem with Heterogeneous Fleet (HMDVRP), among Opt dan Penerapannya”. SKRIPSI Jurusan Matematika-Fakultas
others. MIPA UM. In http://karya-
ilmiah.um.ac.id/index.php/matematika/article/view/24501
[12] Dorigo M. (1992), “Optimization, Learning and Natural
7.REFERENCES Algorithms”, PhD Thesis, Politecnico di Milano, Italia.
[1] Toth P., Vigo D. (2002), The Vehicle Routing problem. Universitá [13] Walter J. Gutjahr. ACO Algorithms with Guaranteed
degli Studi di Bologna, Bologna, Italy. Convergence to the Optimal Solution. Information Processing
[2] Clarke G., Wright J. (1964),"Scheduling of Vehicles from a Letters, 2002, 82: 145-153.
Central Depot to a Number of Delivery Point", Operations
Research, Vol. 12, pp. 568-581.

Metaheuristi ki ACO za Rešavanje Problema Trase Vozila za Više

Skladišta sa Povratnim Utovarom
Jhon Jairo Santa Chávez, Mauricio Granada Echeverri, John Willmer Escobar,
César Augusto Peñuela Meneses

Primljen (11.05.2015.); Recenziran (17.06.2015.); Prihva en (03.07.2015.)

Apstrakt
U ovom radu, metaheuristi ki algoritam zasnovan na optimizaciji kolonije mrava predstavljen je radi
rešavanja problema trase vozila za više skladišta sa dostavom i prikupljanjem pošiljki. Svaka izvršena
trasa sastoji se od jedne pod-trase u kojoj je vožena jedino dostava, pored jedne pod-trase u kojoj je
izvršen jedino proces prikupljanja. Predloženi algoritam pokušava da prona e najbolji redosled posete
klijentima prilikom svake izvedene trase. Pored toga, predloženi pristup odre uje najbolju vezu izme u
pod-trasa dostave i prikupljanja kako bi se dobilo globalno rešenje sa minimumom putnih troškova.
Efikasnost predloženog algoritma ocenjena je posmatranjem niza primera adaptiranih iz literature.
Izra unati rezultati upore eni su sa pohlepnim heuristi kim algoritmom zasnovanim na principu
najbližeg okruženja. Na kraju, predstavljeni su zaklju ci i predlozi za budu i rad.
Klju ne re i: kolonija mrava, povratna pošiljka, kombinatorna optimizacija, izra unata simulacija,
problem trase vozila za više skladišta

IJIEM
 Santa Chávez et al. 57

Table 1. Obtained results over 10 executions with 100 iterations of the proposed algorithm

Greedy Neighbrohood Heuristic Proposed Methodology with ACO

Percetange
Instance
Instance Greedy -
Number Number Computing Number Computing
Average Best Average ACO
of Used Time of Used Time Best Solution
Distance Solution Distance
Vehicles (sec) Vehicles (sec)
1 GJ01H 6 749,8 0,18 719,5 6 693,6 6,92 654,03 9,09%
2 GJ01Q 8 846,8 0,26 808,5 8 777,2 3,55 744,61 7,91%
3 GJ01T 9 826,6 0,27 795,2 9 782,0 7,68 760,45 4,37%
4 GJ02H 3 836,1 0,18 831,9 3 645,4 0,92 598,82 28,02%
5 GJ02Q 4 820,2 0,19 816,3 5 700,2 1,63 685,86 15,98%
6 GJ02T 5 787,5 0,20 770,3 5 665,9 2,45 662,80 13,96%
7 GJ03H 6 1081,1 0,24 1062,0 6 848,6 8,13 827,87 22,04%
8 GJ03Q 8 1066,3 0,26 1012,4 8 923,1 9,96 902,97 10,81%
9 GJ03T 9 996,0 0,26 953,7 9 896,0 6,66 874,41 8,31%
10 GJ04H 8 1422,9 0,41 1416,4 8 1245,7 1,84 1219,61 13,90%
11 GJ04Q 12 1426,0 0,41 1409,9 12 1303,6 4,69 1274,70 9,59%
12 GJ04T 14 1500,5 0,23 1489,7 14 1332,4 6,63 1288,87 13,48%
13 GJ05H 4 1519,2 0,09 1517,1 4 1028,4 0,25 1028,40 32,21%
14 GJ05Q 6 1467,0 0,20 1464,3 6 1136,5 2,34 1069,80 26,94%
15 GJ05T 7 1166,0 0,36 1155,7 7 1049,5 3,16 1027,80 11,07%
16 GJ06H 8 1440,3 0,30 1407,5 8 1155,1 9,05 1112,83 20,93%
17 GJ06Q 12 1445,0 0,44 1408,7 12 1243,0 8,10 1197,10 15,02%
18 GJ06T 14 1406,6 0,71 1373,7 14 1200,1 16,68 1165,34 15,17%
19 GJ07H 8 1342,2 0,29 1306,7 8 1143,9 11,75 1129,42 13,56%
20 GJ07Q 12 1359,4 0,60 1340,6 12 1195,3 18,97 1170,17 12,72%
21 GJ07T 14 1286,1 0,88 1262,4 14 1164,9 11,65 1140,64 9,64%
22 GJ08H 13 7078,1 0,58 6943,6 13 5081,3 24,66 5000,06 27,99%
23 GJ08Q 19 7168,1 0,69 7073,6 19 5951,4 25,29 5808,30 17,89%
24 GJ08T 22 6791,0 0,80 6756,6 22 5874,3 50,19 5779,57 14,46%
25 GJ09H 13 7341,7 0,48 7111,0 13 5208,5 16,92 5164,06 27,38%
26 GJ09Q 19 7257,6 0,67 7036,6 19 6079,8 15,56 5857,96 16,75%
27 GJ09T 22 7229,4 0,54 7117,6 22 6079,0 38,41 5893,50 17,20%
28 GJ10H 13 6830,0 0,57 6404,2 13 4911,6 38,30 4765,45 25,59%
29 GJ10Q 19 7017,2 1,91 6908,1 19 5555,0 128,75 5438,90 21,27%
30 GJ10T 22 6049,2 3,66 5964,2 22 5443,2 112,43 5384,66 9,72%
31 GJ11H 13 6950,8 1,29 6790,2 13 4930,4 84,36 4768,98 29,77%
32 GJ11Q 19 6583,1 3,12 6446,3 19 5299,4 134,72 5193,38 19,44%
33 GJ11T 22 5877,4 7,07 5616,1 22 5123,5 250,39 5039,75 10,26%
Average 3166,4 2564,6 19,01%

IJIEM

View publication stats