Expert Systems with Applications 38 (2011) 1273012738

Contents lists available at ScienceDirect

Expert Systems with Applications

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

A distribution network optimization problem for third party logistics service providers
Hseyin Basligil, Selin Soner Kara , Pelin Alcan, Betl zkan, E. Gzde aglar
Department of Industrial Engineering, Mechanical Faculty, Yildiz Technical University, 34349 Istanbul, Turkey

a r t i c l e

i n f o

a b s t r a c t
Third party logistics service providers (3PLs) have an important role in supply chain management. Increasing cooperation with 3PLs is expanding in todays business environment. Hence, 3PLs need to have an efcient distribution network to meet customer demands. Nevertheless, few researches have tried to propose a solution for distribution network problems of 3PLs. The optimization problem which is discussing in our study is solved in two stages. At the rst stage, the assignment problem which includes assigning the order of the vehicles is solved with mixed integer programming by using GAMS 21.6/CPLEX. The output of the rst stage is used as an input in the second stage. In this stage routes are determined for vehicles by developing a genetic algorithm by using C#. 2011 Elsevier Ltd. All rights reserved.

Keywords: 3PL Distribution network optimization Mixed integer programming Genetic algorithm

1. Introduction Third party logistic (3PL) has become more important for logistic sector in recent years. Companies want to reduce the costs and provide customer satisfaction exactly. They dont want to deal with logistics problems, so they prefer special rms for some or all of their logistics operations. Therefore, a third party logistics (3PL) business is emerging and developing rapidly to fulll the demands for advanced logistics services, in such elds as, transportation, warehousing, freight consolidation and distribution, product marking, labeling and packaging, inventory management, cross docking, product returns, order management, and logistics information systems (Rabinovich, Windle, Dresner, & Corsi, 1999). There are different denitions for 3PL in the literature. Some broad denition for 3PL are the use of external companies to perform logistics functions that have traditionally been performed within an organization (Lieb, 1992) and an external organization that performs all or part of a companys logistics functions (Coyle, Bardi, & Langley, 2003). Bask (2001) describes 3PL as relationships between interfaces in the supply chains and third-party logistics providers, where logistics services are offered, from basic to customized ones, in a shorter or long term relationship, with the aim of effectiveness and efciency. The vehicle routing problem (VRP) can be described as the problem of designing optimal delivery or collection routes from one or several depots to a number of geographically scattered cities or customers, subject to side constraints. The VRP plays a central role
Corresponding author. Tel.: +90 212 383 2902.
E-mail addresses: basligil@yildiz.edu.tr (H. Basligil), ssoner@yildiz.edu.tr (S.S. Kara), palcan@yildiz.edu.tr (P. Alcan), bozkan@yildiz.edu.tr (B. zkan). 0957-4174/$ - see front matter 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.eswa.2011.04.061

in the elds of physical distribution and logistics. The vehicle routing problem lies at the heart of distribution management (Laporte, 1992). A vehicle routing problem (VRP) is one of visiting a set of customers using a eet of vehicles, respecting constraints on the vehicles, customers, drivers, and so on. The goal is to produce a low cost routing plan specifying for each vehicle, the order of the customer visits they make. Industrial VRPs tend to be large, and so local search techniques are used extensively as they scale well and can produce reliably good solutions (Shaw, 1998). Genetic Algorithms (GAs) have seen widespread application to various combinatorial optimization problems, including certain types of vehicle routing problem, especially where time windows are included (Baker & Ayechew, 2003). GAs employ search procedures based on the mechanics of natural selection and survival of the ttest. In the GAs, which use multiple point search instead of single point search and work with the coded structure of variables instead of actual variables themselves, the only information required is the objective function thereby making the searching for global optimum simpler (Das, 2002). Many researchers have conducted studies to solve vehicle routing problems using mathematical models and heuristic algorithms. These are some examples of recent studies: Hadjar and Soumis (2009) used a branch-and-price approach price approach to solve the multiple depot vehicle scheduling problem with time windows (MDVSPTW). They developed a dynamic time window reduction technique. This technique is used at every node of the branchand-price tree to tighten the time window. Chen, Hsueh, and Chang (2009) proposed a nonlinear mathematical model to consider production scheduling and vehicle routing with time windows for perishable food products. They used a constrained NelderMead method and a heuristic method for the vehicle routing with time

H. Basligil et al. / Expert Systems with Applications 38 (2011) 1273012738

12731

window to solve the problem. Zachariadis et al. (2009a) proposed a metaheuristic methodology for the capacitated vehicle routing problem with two-dimensional loading constraints. Tabu search and Guided local search are used together in this methodology. The aim of the paper is to nd the minimum cost routes, starting and terminating at a central depot. Hemmelmayr et al. (2009) developed a new heuristic for the periodic vehicle routing problem without time window. The method is based on variable neighborhood search. In the periodic vehicle routing problem there is a planning horizon of several days and the customers must be visited more than once. Fleszar, Osman, and Hindi (2009) proposed a variable neighborhood search heuristic for open vehicle problem. Proposed solution is based on reversing segments of sub-routes and exchanging segments between routes. Li et al. (2009) proposed a Lagrangian relaxation based-heuristic for the real-time vehicle rerouting problems with time windows. In real-time vehicle rerouting problems there are service disruption because of vehicle breakdowns. So some vehicles must be rerouted. Fuellerer et al. (2009) developed an effective heuristic based on ant colony optimization for the two-dimensional loading vehicle routing problem. There is a combination of two problems: Loading of the freight into the vehicles and routing the vehicles successfully. Zachariadis et al. (2009b) tried to propose an effective hybrid metaheuristic algorithm. The algorithm is a combination of tabu search and guided local search methodologies. There is simultaneous delivery and pick-up service in this type of vehicle routing problem. Genetic Algorithm (GA) is a popular algorithm for solving vehicle routing problems. Some of the studies are as follows: The study of Wang and Lu (2009) primarily focused on solving a capacitated vehicle routing problem (CVRP) by applying a novel hybrid genetic algorithm (HGA) capable of practical use for manufacturers. The search mechanism embedded in the GA focuses on the breeding process in evolution on crossover and mutation operators that are applied using probability setting to approach a close-to-optimal solution. Liu et al. (2009) studied the eet size and mix vehicle routing problem (FSMVRP), in which the eet is heterogeneous and its composition to be determined. They design and implement a genetic algorithm (GA) based heuristic. On a set of twenty benchmark problems it reaches the best-known solution 14 times and nds one new best solution. It also provides a competitive performance in terms of average solution. The paper of Prins (2009) presented two memetic algorithms (genetic algorithms hybridized with a local search) able to solve both the VFMP (The vehicle eet mix problem) and the HVRP (The heterogeneous eet VRP). They are based on chromosomes encoded as giant tours, without trip delimiters, and on an optimal evaluation procedure which splits these tours into feasible trips and assigns vehicles to them. Ho et al. (2008) studied the MDVRP (multi-depot vehicle routing problem) because the number of depots is not limited to one in many real-world situations. Besides routing and scheduling, the grouping problem is also considered in the MDVRP. Because the MDVRP integrates three hard optimization problems, a Hybrid genetic algorithm (HGA) rather than a simple GA was developed. The paper of Prins (2004) bridged the gap by presenting a relatively simple but effective hybrid GA. The framework of this research is the development of effective metaheuristics for hard combinatorial optimization problems met in vehicle routing. It is surprising to notice in the literature the absence of effective genetic algorithms (GA) for the vehicle routing problem contrary to node routing problems with time windows or arc routing problems. The study of Baker and Ayechew (2003) considers the application of a genetic algorithm (GA) to the basic vehicle routing problem (VRP), in which customers of known demand are supplied from a single depot. There are limited numbers of papers about vehicle routing problems of 3PLs. Some examples can be given as follows: Krajewska and Kopfer (2009) used a tabu search algorithm for

solving the integrated transportation planning problem. Zpfel and Bgl (2008) consider short-range weekly planning on the part of postal companies that must decide about pickup tours and delivery tours for uctuating volume (number of shipments), with time windows for the demand points, in consideration of variable vehicle capacities and personnel planning, and including outsourcing decisions for tours and drivers for 3PLs. Tan et al. (2006) developed a hybrid multi-objective evolutionary algorithm for truck and trailer vehicle routing problem. The purpose of this study is to minimize the routing distance and the number of trucks. The remainder of this paper is organized as follows. The theoretic descriptions for genetic algorithms are presented in Section 2. The proposed methodology is described in Section 3. Section 3 also, explains a numerical example in Istanbul. Finally, the results and the conclusion are presented in Section 4.

2. Genetic Algorithm Genetic Algorithms (GAs) are adaptive heuristic search algorithm premised on the evolutionary ideas of natural selection and genetic (De Jong, Spears, & Gordon, 1993). Genetic Algorithms are a family of computational models inspired by evolution. These algorithms encode a potential solution to a specic problem on a simple chromosome like data structure and apply recombination operators to these structures so as to preserve critical information (Whitley, 1994). The basic concept of GAs is designed to simulate processes in natural system necessary for evolution. As such they represent an intelligent exploitation of a random search within a dened search space to solve a problem. First pioneered by John Holland in the 60s, Genetic Algorithms has been widely studied, experimented and applied in many elds in engineering worlds. Not only does GAs provide an alternative method to solve problem, it consistently outperforms other traditional methods in most of the problems link (De Jong et al., 1993). Genetic algorithms have shown great advantages in solving the combinatorial optimization problem in view of its characteristic that has high efciency and that is t for practical application (Chiu, Fang, & Lee, 1999). Genetic Algorithms are search algorithms which are based on natural selection and natural genetic mechanism (Cheng, 1999). From the view point of the working principle, genetic algorithms rstly needs the coding of the problem with the condition that it should be tting with the GA (akar, Koker, & Demir, 2008). After coding stage, GA operators are applied on chromosomes. It is not guaranteed that the obtained new offsprings are good solutions by the working of crossover and mutation operators. Feasible solutions are evaluated, and others are left out of evaluation. The feasible ones of the obtained offsprings are taken and new populations are formed by reproduction process using these offsprings (Damodaran, 2006). During each iteration step genetic operations, that is, crossover, mutation and natural selection are applied in order to search potential better solutions. Crossover combines two chromosomes to generate next-generation chromosomes preserving their characteristics. Mutation reorganizes the structure of genes in a chromosome randomly so that a new combination of genes may appear in the next generation. Reproduction is to copy a chromosome to the next generation directly so that chromosomes from various generations could cooperate in the evolution and the quality of the population may be improved after each generation (Hop & Nagarur, 2004). An implementation of a genetic algorithm begins with a population of (typically random) chromosomes. One then evaluates these structures and allocates reproductive opportunities in such a way that those chromosomes which represent a better solution

12732

H. Basligil et al. / Expert Systems with Applications 38 (2011) 1273012738

Begin

Steps of the methodology are as follows:  Pre-research stage: Determining the transportation cost.  Stage 1: Routing the incoming requests according to vehicle types.  Stage 2: Determining the vehicle routes. 3.1.1. Pre-research stage: Determining the transportation cost For transporting the loads two types of vehicles are intended in the model. These are trucks and trailers. While truck is carrying up to 20 tons, trailers are carrying up to 30 tons as an acceptance. The cost items of a logistic rm can be listed as follows (Dondo & Cerd, 2009): 3.1.1.1. Fixed costs: Driver cost, amortization, insurance and taxes. Fixed costs occur independently from the trip. Variable costs are proportional with the vehicle miles. Factors affecting the driver cost can be listed as driver salary, allowance, road, service, social rights, and education costs. The daily cost value can be calculated with all these costs. Amortization is the process of increasing or accounting for, an amount over a period of time (Liu, Huang, & Ma, 2009). Consequently amortization values are calculated for each vehicle types. Annual insurance and tax expenditures belonging to vehicles are occurring in the xed costs. These are motor vehicles tax, automobile insurance, trafc insurance, liability insurance carrier, examination fee insurance and tax costs. 3.1.1.2. Variable costs: Fuel cost, maintenance cost. Fuel cost constitutes the largest part of the transportation cost. Fuel cost per km is calculated (unit cost of fuel burning rates). Also, road and weather conditions can affect the burning rate. In this article, up to 26 tons of the transport, the burning rate is taken as 34.5% in the trailers vehicles. But it has been calculated as 36.5% for carrying more than 26 tons. When the vehicles reach specic kilometer values, maintenance must be done to the vehicles. With all these costs are taken into consideration, the cost of providing transportation functions are computed with obtained data from a Turkish Logistics Firm for each vehicle type. The results are shown in Table 1. As it can be seen from the table, variable cost coefcient is 0.88 for the loads up to 26 tons in the trailers vehicle. But this coefcient is 0.92 for more than 26 tons. This difference is caused by fuel cost. So the trailers vehicles can be divided into two separate types. Thus in this study three types of vehicles are included in carrying loads. These types can be seen in Table 2 with their capacities. 3.1.2. Stage 1: Routing the incoming requests according to vehicle types A mixed integer programming model is built in the rst stage of the proposed methodology to assign incoming orders to the vehicles. GAMS 21.6/CPLEX is used for the solution of this assignment problem. The notation can be represented as follows: Subscripts i Departure location i = 1, 2, 3 j Arrival location j = 1, 2, 3 k Vehicle type k = 0, 1, 2 Parameters Kk Maximum vehicle capacity for vehicle type k SMk Fixed cost for for vehicle type k DMk Variable cost for vehicle type k Mij The distance from point i to point j Sij Order quantity from point i to point j Variables xijk The number of trips from point i to point j with vehicle type k

Generating an initial population Next Generation

Evaluation/ Fitness Computing

Mutation
Crossover

No
Reproduction Stop

End Yes
Fig. 1. Genetic Algorithm.

to the target problem are given more chances to reproduce than those chromosomes which are poorer solutions. The goodness of a solution is typically dened with respect to the current population (Whitley, 1994). The rst step in the implementation of any genetic algorithm is to generate an initial population (Holland, 1975). In most cases the initial population is generated randomly. After creating an initial population, each string is then evaluated and assigned a tness value. It is helpful to view the execution of the genetic algorithm as a two stage process. It starts with the current population. Selection is applied to the current population to create an intermediate population. Then recombination and mutation are applied to the intermediate population to create the next population. The process of going from the current population to the next population constitutes one generation in the execution of a genetic algorithm (Goldberg, 1989). Crossover is a genetic operator that combines (mates) two chromosomes (parents) to produce a new chromosome (offspring). The idea behind crossover is that the new chromosome may be better than both of the parents if it takes the best characteristics from each of the parents. In genetic algorithms, mutation is a genetic operator used to maintain genetic diversity from one generation of a population of chromosomes to the next (Muhuri & Shukla, 2008). It is analogous to biological mutation. Mutation operator creates new chromosomes by causing small perturbations in genes. This operation prevents from getting stuck on local suboptimal solutions and its very helpful to maintain the richness of the population in dealing with large scale problems (Chiu et al., 1999). After the process of selection, recombination and mutation is complete, the next population can be evaluated. The process of evaluation, selection, recombination and mutation forms one generation in the execution of a genetic algorithm (Whitley, 1994). Fig. 1 summarizes the concept of GA.

3. A numerical example 3.1. The proposed methodology In the pre-research phase the transportation costs of vehicles are determined. Our methodology is solved in two stages. In the rst stage, incoming requests are routed according to vehicle types. GAMS 21.6/CPLEX software is used at the rst stage. The output of the rst stage is used as an input in the second stage for determining the vehicle routes. Besides, in the second level, C# programming language and a genetic algorithm software have been developed for the vehicle routes. Fig. 2 depicts the proposed optimization model.

H. Basligil et al. / Expert Systems with Applications 38 (2011) 1273012738

12733

1.VEHICLE TYPES 2.COSTS 3.DISTANCES

ORDER LST

GAMS
MIXED INTEGER PROGRAMING

ORDER LIST TRIP LIST

C# SOFTWARE
GENETIC ALGORITHM

OPTIMAL ROUTES

LOCATIONS AND QUANTITIES OF OWN VEHICLES

COSTS OF SPOT VEHICLES BASED ON ROUTES AND VEHICLE TYPES

Fig. 2. Optimization model.

Table 1 The cost functions according to the vehicle types. Cost function = A + BX Truck A B Trailers A B 026 ton 108.66 0.62 159.91 0.88 +26 ton 0.92

min

XXX
i j k

xijk SM k DMk M ij

s:t:

P
k

K k xijk P Sij xijk 6 Sij

8i; j

2 3 4

K0 K2

P
k

8i; j

P
k

xijk P Sij 0 8i; j

Constraint (1) is the objective function that minimizes the cost of trip, Constraint (2) guarantees that all orders are satied, Constraint (3) and Constraint (4) ensure the maximum and the minimum numbers for the vehicles that are used for trip respectively. 3.1.6. Stage 2: Determining the vehicle routes In the second phase of the model, the solution of vehicle routing problem is obtained with the help of a meta-heuristic, genetic algorithm. The problem is solved with a programming language which is written in Microsoft Visual C# 2005 Express Edition. The program automatically receives the information of the order and trip lists. Then, it prot from the data to determine the optimal vehicle route. Genetic algorithm is solved separately for truck and

trailer. For the trucks the routes are combined in a proper way to determine optimal routes. So, that the routes are created easily. Also, the same procedure is applied for trailers. First, trip from departure-arrival points which are determined in the rst stage are numbered individually. The data which is numbered is designated as one route. The route consists of departure point, arrival point and vehicle type information. The purpose is creating a permutation encoding that constitutes the chromosome series in the solution of the genetic algorithm. A coding structure which consists of 01 chromosomes is created for the division of the route series. According to GAMS output six trips have be done by using truck. For these six trips the route numbers will be assigned to the routes in the program in Table 3. Hence, a (1, 0, 2, 4, 3, 5) sequence is a route sequence. This route series is divided into the routes. In the section part, 0 represents that the route continue but 1 represents the end of the route. Then, there is chromosome logic occur which is shown in Fig. 3: Accordingly, such as the above structure of the chromosomes shows us two trucks will be used. In this way, while the rst truck is routing to (1, 0, 2, 4) points, second truck routes to {3, 5} points respectively. According to the working principle of genetic algorithm, a random sequence of the chromosome is created primarily for the rst generation with the number of determined generation. After that, for creating the other generations, individual elitism, crossover and mutation operations are used in the current generation. In elitism, the individuals of the generation are sorted according to cost value in descending order. Then, the elitism rate generation size unit individuals of the rst generation
Table 2 Used vehicle types and capacities. Vehicle types Truck Trailer 1 Trailer 2 Max capacity 20 ton 26 ton 30 ton

12734 Table 3 Example route numbering. Trip number 0 1 2 3 4 5 Departure 1 1 1 2 2 3

H. Basligil et al. / Expert Systems with Applications 38 (2011) 1273012738

Arrival 2 2 3 3 3 1

ARRAY 1: ROUTES

ARRAY 2: SECTIONS

Fig. 3. Chromosome structure.

transferred to the next generation starting from the individual with minimum cost Fig. 3. On the other hand, crossover rate generation size unit individuals are being formed after the crossover process. In our methodology, tournament method is used for chromosome sequence selection. In the rst series of chromosome consists of permutation encoding, a common cut point for parents is determined randomly. The new chromosome series contain the part of parents chromosome till cut point. After the part of the cut point, the routes are listed according to the other parents chromosome series. This crossover is showed in Fig. 4. Also, standard crossover operation is performed for the second series of 01 coded chromosome. This crossover is showed in Fig. 5. The rest of the mutation rate generation size unit individuals are created by randomly selected individuals. They are mutated from the previous generation. In the mutation process, two random numbers that are selected for the rst sequence of chromosome are replaced. For the second sequence of chromosome, number 0 is converted to number 1 and also number 1 is transformed to number 0 in two points which are selected randomly. Fig. 6 shows a mutation example. After the iteration number of generation creating reaches to the maximum iteration number, individual of the last generation which has minimum cost is the solution of the optimization problem. The cost results are used for owned vehicles in calculating an individuals tness function. Daily list prices are used which are designated between the routes for spot vehicles. In addition to
PARENT 1

these costs, a cost caused by time constraints of the routes is taken into account as a penalty cost. Due to legal obligations, vehicle drivers do not exceed 9 h as driving time of 24-hour time period. For this reason, vehicle routes have be done in a 9 h period. If the duration of a route is more than 9 h, this condition is located in tness function as a penalty cost. The duration of trip depends on many internal and external factors such as distance, speed, road conditions, weather conditions and trafc conditions. To reect this uncertainty to the model, probability has been used. Maximum speed limit is 80 km/s on interurban roads but is 90 km/s on highways for trucks and trailers. Except in extraordinary cases, on intercity highways minimum speed limit is 15 km/s but on highways it is 40 km/s. With the help of this information, the speed of vehicles is assumed to represent a triangular distribution. Moreover, in triangular distribution logic, the speed of vehicles are regarded as min. 35 km/s, average 65 km/s and max. 90 km/s. By dividing the distance to these three speed value, a time period of triangular structure is obtained. (a, b, c) shows the completion time of a route with min. a time, average b time and max. c time. While considering 9-hours limit, four different conditions can be arisen. Fig. 7 shows four different conditions.In Condition (1) (if c 6 9), penalty cost does not occur because time limit is not exceeded. This is the most ideal situation. In Condition (2) (if b 6 9 < c) the blue painted area represents the unwanted area. The probability of exceeding 9 h is calculated. However average time does not exceed 9 h so this condition can be assumed as an acceptable solution. By multiplying the probability and penalty coefcient, penalty cost is calculated. In Condition (3) (if a 6 9 < b), an unwanted situation is being occurred. The probability of exceeding 9 h is high. By multiplying the probability and penalty coefcient, penalty cost is created. In Condition (4) (if a P 9) is not wanted like as previous one. For this reason, an extreme high cost is added as a penalty cost for such a solution. The sum of vehicle cost and penalty cost is considered to be a tness value for each individual. Thereby comparison can be done between individuals. Thus, the lowest-cost individual is the most appropriate individual. The algorithm of the methodology is depicted in Fig. 8.

3.2. A numerical example A numerical example is solved to represent the efciency of our methodology. The model is used for trying to nd the number of trips with which vechicle type. The orders in the receiving and the delivery points are collected as cumulative. In the problem, trip of the products is carried out in two ways. One option is to use the companys own vehicles. They will be assigned to trip and then if
PARENT 2

CHILD 1 TAKES DIRECTLY

LINE UP ACCORDING TO PARENT 2

CHLD 2 TAKES DIRECTLY

LINE UP ACCORDING TO PARENT 1

CHILD 1

CHILD 2

Fig. 4. Crossover permutation method for the rst sequence of chromosome.

H. Basligil et al. / Expert Systems with Applications 38 (2011) 1273012738

12735

PARENT 1

PARENT 2

FIRST PART OF CHILD1

REST PART OF CHILD 2

FIRST PART OF CHILD2

REST PART OF CHILD1

CHILD 1

CHILD 2
Fig. 5. Crossover method for the second series of chromosome.

The assumptions of the problem are as follows:

CHROMOSOME 2 8 6 4 7 1 3 5

1. The loadings are represented in ton not volume. 2. The moving materials are uniform. 3. There is no limitation for shipping different materials at the same time. Two types of vehicles are used in this study rst type is the vehicles that rm owns and the second type is spot vehicles that rm rents. In all three locations one truck and two trailers exist. Spot vehicle prices are 200 TL and 400 TL for trucks and truck trailers respectively. Table 5 shows the mixed integer programming results. According to this, 6 routes with truck and 17 routes with trailer can be used for trip. The genetic algorithm parameters are as follows:      Generation Size: 200. Max. iteration number: 100. Elitism Ratio: 0.1. Crossover Ratio: 0.8. Mutation Ratio: 0.1.

CHROMOSOME FORMED AFTER MUTATION

Fig. 6. Mutation method for the rst sequence of chromosome.

there is a still need in the process, another option is to rent vehicles from the spot market. An order list is created on Excel Programming. In Table 4, the orders are available between three different locations. Material quantities which should be moved between these locations are shown in the ordered list.

DURATION

9
DURATION

a
Condition 1

Condition 2

DURATION

9 a

Condition 3

Condition 4

Fig. 7. Possible conditions.

12736

H. Basligil et al. / Expert Systems with Applications 38 (2011) 1273012738

Fig. 8. The algorithm of the methodology.

The results obtained from mixed integer programming model are transferred to an access database le. In this le, created order list and the trip list which is obtained from the mixed-integer programming appeared. The result of this program is realized that how many owned and spot vehicles have to be used to satisfy the orders. Additionally, all the costs with the including and excluding penalty cost can be seen from the program output.

The algorithm has been operated in 100 times. Best solutions output is as follows, one owned truck, one spot truck, six owned trailers will be used for trip. Out of the penalty cost, the cost is 668.26 TL and 3286.66 TL for trucks and trailers respectively. The total cost is 3954.92 TL to fulll the orders. In the optimization nal report, it is seen that there will be 580 km road which will be covered by the owned truck. Besides, it is easily said that

H. Basligil et al. / Expert Systems with Applications 38 (2011) 1273012738 Table 4 Order list. Order list (Ton) 1 1 2 3 150 46 2 100 24 3 200 99 Table 7 Best solution for trucks. Truck route data Route 0 Vehicle supply class Total Km Optimist time Average time Pessimist time Cost (with time coefcient) Cost Trips From location 1 to location 2 From location 2 to location 3 From location 3 to location 1 From location 1 to location 2 From location 2 to location 3 Owned 580 6.44 8.92 16.57 548,68 468.26 20 ton 20 ton 20 ton 20 ton 20 ton Route 1 Spot 200

12737

Table 5 The results. Departure 1 1 1 1 2 2 2 3 3 3 Arrival 2 2 3 3 1 3 3 1 1 2 Vehicle type Truck Trailer-2 Truck Trailer-2 Trailer-2 Truck Trailer-2 Truck Trailer-1 Trailer-1 The number of trip 2 2 1 6 5 2 2 1 1 1

200 Trips From location 1 to location 3

20 ton

this distance will be covered respectively in 6.44 h, 16.57 h and 8.92 h in the best case scenario, the worst-case scenario and average condition. The truck will be started from the rst location and will be nished at the third location. Trip for owned truck is 1-2-3-1-2-3 locations with the 20 ton load. Also, Tables 6 and 7 summarize the best solution results for trailers and trucks, respectively.

have been met substantially with the aim of low-cost route suggestions. Mixed integer programming has been used for the assignment of the orders to the vehicles with the different capacities. Subsequently the suitable routes have been determined with the help of genetic algorithm method. It is seen that, this solution is such a considerably acceptable solution with the low cost. At the same time, it is seen that nding the solution is getting hard if the size of the problem increases. But this problem is deliberated; the optimal solution can be useful for todays logistics companies.

References
Baker, B. M., & Ayechew, M. A. (2003). A genetic algorithm for the vehicle routing problem. Computers & Operations Research, 30, 787800. Bask, A. H. (2001). Relationships among TPL providers and members of supply chainsA strategic perspective. Journal of Business & Industrial Marketing, 16(6), 470486. Chen, H., Hsueh, C., & Chang, M. (2009). Production scheduling and vehicle routing with time windows for perishable food products. Computers & Operations Research, 36, 23112319.

4. Results and conclusion In this study, a new methodology is developed to determine optimal routes for vehicles with different capacities and different costs for logistic companies. In the model, the customer orders

Table 6 Best solution for trailers. Route 0 Trailer route data Vehicle supply class Total Km Optimist time Average time Pessimist time Cost (with time coefcient) Cost Trips From location 1 to location 3 From location 3 to location 2 From location 2 to location 1 From location 1 to location 2 From location 2 to location 3 Route 3 Vehicle Supply Class Total Km Optimist Time Average Time Pessimist Time Cost (with time coefcient) Cost Trips From location 2 to location 3 From location 3 to location 2 From location 2 to location 1 From location 1 to location 3 Owned 580 6.44 8.92 16.57 808.26 689.91 30 ton Free trip 30 ton 30 ton 30 ton Owned 480 5.33 7.38 13.71 664.9 597.91 29 ton Free trip 30 ton 30 ton Route 1 Owned 400 4.44 6.15 11.43 553.51 527.91 Trips From location 1 to location 2 From location 2 to location 1 From location 1 to location 3 Route 2 Owned 390 4.33 6 11.14 536.08 515.11 Trips From location 3 to location 2 From location 2 to location 1 From location 1 to location 3

30 ton 30 ton 30 ton

24 ton 30 ton 30 ton

Route 4 Owned 400 4.44 6.15 11.43 545.51 519.91 Trips From location 3 to location 1 From location 1 to location 3

Route 5 Owned 300 3.33 4.62 8.57 435.91 435.91 Trips From location 2 to location 1 From location 1 to location 3

26 ton 30 ton

30 ton 30 ton

12738

H. Basligil et al. / Expert Systems with Applications 38 (2011) 1273012738 Hop, N. V., & Nagarur, N. N. (2004). The scheduling problem of PCBs for multiple non-identical parallel machines. European Journal of Operational Research, 158, 577594. Krajewska, M. A., & Kopfer, H. (2009). Transportation planning in freight forwarding companies-Tabu search algorithm for the integrated operational transportation planning problem. European Journal of Operational Research, 197, 741751. Laporte, G. (1992). The vehicle routing problem: An overview of exact and approximate algorithms. European Journal of Operational Research, 345, 358. Li, J., Mirchandani, P. B., & Borenstein, D. (2009). Real-time vehicle rerouting problems with time windows. European Journal of Operational Research, 194, 711727. Lieb, R. C. (1992). The use of third-party logistics services by large American manufacturers. Journal of Business Logistics, 13(2), 2942. Liu, S., Huang, W., & Ma, H. (2009). An effective genetic algorithm for the eet size and mix vehicle routing problems. Transportation Research, 45(3), 434445. Muhuri, P. K., & Shukla, K. K. (2008). Real-time task scheduling with fuzzy uncertainty in processing times and deadlines. Applied Soft Computing, 8, 113. Prins, C. (2004). A simple and effective evolutionary algorithm for the vehicle routing problem. Computers & Operations Research, 31, 19852002. Prins, C. (2009). Two memetic algorithms for heterogeneous eet vehicle routing problems. Engineering Applications of Articial Intelligence, 22(6), 916928. Rabinovich, E., Windle, R., Dresner, M., & Corsi, T. (1999). Outsourcing of integrated logistics functions. International Journal of Physical Distribution and Logistics Management, 29(6), 353373. Shaw,P. (1998). Implementation of LNS for constrained VRPs. APES Report. http:// www.cs.strath.ac.uk/~apes. Tan, K. C., Chew, Y. H., & Lee, L. H. (2006). A hybrid multi-objective evolutionary algorithm for solving truck and trailer vehicle routing problems. European Journal of Operational Research, 172, 855885. Wang, C. H., & Lu, J. Z. (2009). A hybrid genetic algorithm that optimizes capacitated vehicle routing problems. Expert Systems with Applications, 36, 29212936. Whitley, D. (1994). A genetic algorithm tutorial. Statistics and Computing, 4, 6585. Zachariadis, E. E., Tarantilis, C. D., & Kiranoudis, C. T. (2009a). A guided tabu search for the vehicle routing problem with two-dimensional loading constraints. European Journal of Operational Research, 195, 729743. Zachariadis, E. E., Tarantilis, C. D., & Kiranoudis, C. T. (2009b). A hybrid metaheuristic algorithm for the vehicle routing problem with simultaneous delivery and pick-up service. Expert Systems with Applications, 36, 10701081. Zpfel, G., & Bgl, M. (2008). Multi-period vehicle routing and crew scheduling with outsourcing options. International Journal of Production Economics, 113(2), 980996.

Cheng, L. M. W. (1999). A genetic algorithm for minimizing the makespan in the case of scheduling identical parallel machines. Articial Intelligence in Engineering, 13, 399403. Chiu, N. C., Fang, S. C., & Lee, Y. S. (1999). Sequencing parallel machining operations by genetic algorithms. Computers & Industrial Engineering, 36, 259280. Coyle, J. J., Bardi, E. J., & Langley, C. J. (2003). The management of business logisticsA supply chain perspective. Mason: South- Western Publishing. akar, T., Koker, R., & Demir, H. I. (2008). Parallel robot scheduling to minimize mean tardiness with precedence constraints using a genetic algorithm. Advances in Engineering Software, 39, 4754. Damodaran, P., Manjeshwar, P. K., & Srihari, K. (2006). Minimizing makespan on a batch-processing machine with non-identical job sizes using genetic algorithms. International Journal of Production Economics, 103, 882891. Das, D. (2002). Reactive power compensation for radial distribution networks using genetic algorithm. Electrical Power and Energy Systems, 24, 573581. De Jong, K. A., Spears, W. M., & Gordon, D. F. (1993). Using genetic algorithms for concept learning. Machine Learning, 13, 161188. Dondo, R. G., & Cerd, J. (2009). A hybrid local improvement algorithm for largescale multi-depot vehicle routing problems with time windows. Computers & Chemical Engineering, 33, 513530. Fleszar, K., Osman, I. H., & Hindi, K. S. (2009). A variable neighborhood search algorithm for the open vehicle routing problem. European Journal of Operational Research, 195, 803809. Fuellerer, G., Doerner, K. F., Hartl, R. F., & Iori, M. (2009). Ant colony optimization for the two-dimensional loading vehicle routing problem. Computers & Operations Research, 36, 655673. Goldberg, D. (1989). Genetic algorithms in search. Optimization and Machine Learning. Reading, MA: Addison Wesley. Hadjar, A., & Soumis, F. (2009). Dynamic window reduction for the multiple depot vehicle scheduling problem with time windows. Computers & Operations Research, 36, 21602172. Hemmelmayr, V. C., Doerner, K. F., & Hartl, R. F. (2009). A variable neighborhood search heuristic for periodic routing problems. European Journal of Operational Research, 195, 791802. Ho, W., Ho, G. T. S., Ji, P., & Lau, H. C. W. (2008). A hybrid genetic algorithm for the multi-depot vehicle routing problem. Engineering Applications of Articial Intelligence, 21, 548557. Holland, J. (1975). Adaptation in natural and articial systems. University of Michigan Press.