International Journal of Trend in Research and Development, Volume 3(2), ISSN: 2394-9333

www.ijtrd.com

Mixed Integer Programming Model for Split

Delivery Vehicle Routing Problem with Fleet and
Driver Scheduling
Arnita Herman Mawengkang
State University of Medan, Department of Mathematics
Medan, Indonesia University of Sumatera Utara, Medan, Indonesia

Abstract: Vehicle Routing Problem (VRP) is a key element this paper it is assumed that the capacity of each vehicle used
of many transportation systems which involve routing of is homogeny. A survey paper regarding to SDVRP is
fleet of vehicles from a depot to a set of customers node. It is presented by [10].
required that these vehicles return to the depot after serving
Firstly, the SDVRP as the variant of VRP was introduced
customers’ demand. This paper investigates a relaxed version
by [8]. They also gave some properties and a local heuristic
of VRP, in which the number of visits to the customer is not
search for solving the problem. Due to the structure of the
restricted to be at most one. The relaxed version is called split
problem, there are two kind of solution approaches have been
delivery VRP. The problem incorporates time windows, fleet
proposed ,i.e., exact and heuristic methods.
and driver scheduling in the planning horizon. The goal is to
schedule the deliveries according to feasible combinations of The first exact method for solving SDVRP was addressed
delivery days and to determine the scheduling of fleet and by [9]. They introduced a mathematical formulation based on
driver and routing policies of the vehicles. The objective is to integer programming and solved through a cutting plane
minimize the total costs of all routes over the planning approach.[34] and [28] proposed the problem with time
horizon. We model the problem as a linear mixed integer windows and present exact approaches based on column
program. We develop a combination of heuristics and exact generation and branch-and-bound techniques. A column
method for solving the model. generation approach was presented by [15]. Columns include
route and delivery amount information. They solved the
Keywords: Routing, Scheduling, Transportation, Combined method pricing sub-problems by a limited-search-with-bound
algorithm. Feasible solutions are obtained iteratively by fixing
I. INTRODUCTION
one route once. [21] used branch and price method for the
In logistic system it is necessarily to design the optimal SDVRP formulated as a mixed integer program based on arc-
routes for a set of vehicle involved with given capacity in such flow consideration. A branch-price-and-cut algorithm was
a way that could satisfy customers’ demand. Vehicle Routing proposed by [24] for solving commodity constrained of
Problem (VRP) is one of the important tool that can be used SDVRP. They formulated the problem through a set
for designing the optimal routes. This well known partitioning pattern. [23] present two exact branch-and-cut
combinatorial optimization problem consists of a customer solution methodologies.
node with deterministic demands, and a central depot which
acts as the base of a fleet of vehicles. The customer’s demand [32] develop a column generation technique to address the
must be delivered by exactly one vehicle The objective is to problem of scheduling helicopter flights to exchange crews on
design a set of Hamiltonian cycles (vehicle routes) starting off-shore platforms. They model the problem as an SDVRP
and terminating at the central depot, such that the demand of and propose an integer linear programming formulation in
customers is totally satisfied, the total demand of the which all feasible flight schedules are enumerated in advance,
customers assigned to a route does not exceed vehicle and solve its linear relaxation by means of column generation.
capacity, such that minimizing the overall travel cost, taking [13] proposed a dynamic program with finite state and action
into account various operational constraints. VRP was first spaces, and tackle it by solving the shortest path problem on a
introduced by [5]. A comprehensive overview of the Vehicle digraph whose nodes and arcs correspond to states and
Routing Problem can be found in [4] which discusses transitions between states, respectively. [14] considered the
problem formulations, solution techniques, important variants undirected version of the problem and presented new lower
and applications. There are many other researchers have been bounds together with some polyhedral results for the SDVRP.
working in this area to discover new methodologies. A An integer programming model was introduced and a
comprehensive interesting survey of the Vehicle Routing relaxation of the SDVRP. They showed that all constraints in
Problem can be found in [4], [3], [36], [37], and [7]. [6] this relaxation are facet-defining for the convex hull of the
addressed a thorough review of past and recent developments incidence vectors of the SDVRP solutions.
of VRP. [8] developed a heuristic method for SDVRP, which
In some cases, particularly, when a customer’s demand involved a two-stage local search algorithm. Their method
exceeds the vehicle capacity it is necessary to visit that was based on the VRP route improvement procedure of [29]
customer more than once, it requires only a little more thought The first stage is to find the solution of VRP problem, and
to see that even when all customer demands are less than or then an SDVRP solution is constructed and improved in the
equal to the vehicle capacity, it may be beneficial to use more second stage. Other heuristic approaches were developed
than one vehicle to serve a customer. For this situation it is using hybrid methods and metaheuristics. [11] and [20]
necessarily to have split delivery In terms of VRP this type of proposed a tabu search algorithm, a memetic algorithm with
problem is called the split delivery vehicle routing problem population management by [26], three hybrid algorithms due
(SDVRP), where the single-visit assumption is relaxed and to [16], [12], and [30], a metaheuristic based on the scatter
each customer may be served by more than one vehicle. In search methodology by [31], an adaptive memory search-

IJTRD | Mar-Apr 2016

Available [email protected] 250
 International Journal of Trend in Research and Development, Volume 3(2), ISSN: 2394-9333
www.ijtrd.com
based metaheuristic by [18], a population-based tabu search respectively. We assume the number of vehicles equals to the
with vocabulary building approach by [17], a construction number of drivers. Denote the set of n customers (/nodes) by
heuristic due to [33], a randomized granular tabu search N  1, 2,, n . Denote the depot by 0, n  1 . Each
technique by [25]. [22] used the combination of local search
and ant colony optimization. A variable neighborhood descent vehicle starts from 0 and terminates at n  1 . Each
metaheuristic was due to [35]. For a more detailed discussion
on heuristics can be found in [19]. customer i  N specifies a set of days to be visited, denoted
This paper concerns with a comprehensive model for the by Ti  T . On each day t  Ti , customer i  N requests
SDVRP incorporated with time windows, fleet and driver service with demand of qit in weight and pit in volume,
scheduling (SDVRPFDTWP) . The basic framework of the
vehicle routing part can be viewed as a Heterogeneous split service duration d it and time window  ai , bi  . Note that, for
delivery Vehicle Routing Problem with Time Windows
(HSDVRPTW) in which a limited number of heterogeneous the depot i 0, n  1 on day t, we set qit  pit  dit  0 .
vehicles, characterized by different capacities are available
and the customers have a specified time windows for services.
The travel time between customer i and j is given by cij .
We propose a mixed integer programming formulation to Denote the cost coefficients of the travel time of the drivers by
model the problem. A feasible neighborhood heuristic search A.
based on active set constraint is addressed to get the integer t
We define binary variable xijk to be 1 if vehicle k travels
feasible solution after solving the continuous model of the
problem. from node i to j on day t, Variable vikt is the time that vehicle
II. MATHEMATICAL FORMULATION k visits node i on day t. Binary variable zikt indicates whether
Using graph, SDVRP can be defined as follows. Let G = vehicle k takes a break after serving customer i on day t.
(V, E) be a complete graph, where V = {v0, v1, …, vn} is the
Variable uikt is the elapsed driving time for vehicle k at
vertex set and E = {(vi, vj): vi, vj∈V, i ≠ j, j ≠ 0} is the set of
traversing route. Vertex v0 represents the central depot customer i after the previous break on day t. Binary variable
where a fleet of vehicles is located. The vehicles are ylkt is set to 1 if vehicle k is assigned to driver l on day t.
assumed to be identical with maximum carrying load equal
to Q. The remaining n vertices of V \ {v0} represent the Variables rlt and slt are the total working duration and the
customer set. Each customer vertex is associated a non- total travel time for driver l on day t, respectively.
negative known demand qi, whereas with each arc (vi, vj) ∈ E
is associated a cost cij which corresponds to the cost (travel This notations used are given as follows :
time, distance) for traversing from vi to vj. Also, it is Set:
assumed that the triangle inequality holds. The delivery T The set of workdays in the planning horizon,
services for each vehicle must start and end its D The set of drivers,
traversing route at the depot, v0. Tl The set of workdays for driver l ∈ D,
The demand of a customer must be satisfied and may K The set of vehicles,
be fulfilled by more than one vehicle. In order to meet a N The set of customers,
customer’s demand split delivery is needed. One of the N0 The set of customers and depot N0 = {0, n + 1}
reason this condition occur is that customer’s demand ∪ N,
exceeds vehicle capacity. Mathematically, it can be Ti The set of days on which customer i  N
written as Q  qi . orders,

The objective of the problem is to design the set of Parameter:

Hamiltonian paths to serve all customers such that: the number Qk The weight capacity of vehicle k  K ,
of vehicles used is minimized, as well as to minimize the total Pk The volume capacity of vehicle k  K ,
distance or cost of the generated paths. There are some
restrictions which must be satisfied, such as, every path cij The travel time from node i  N0 to node
originates from the central depot v0, each customer vertex is j  N0 ,
assigned to a single path, and the total demand of the customer
[ai, bi] The earliest and the latest visit time at node
set assigned to a single path does not exceed the maximum
carrying load Q of the vehicles (capacity constraint). i  N0 ,
To formulate the model, firstly we denote T as the d it The service time of node i  N0 on day
planning horizon and D as the set of drivers. The set of
t  Ti ,
workdays for driver l  D is denoted by Tl  T . The start
working time and latest ending time for driver l  D on day
q t
i The weight demand of node i  N0 on day

t  T are given by g lt and hlt , respectively. For each driver t  Ti ,

l  D , let H denote the maximum weekly working duration. pit The volume demand of node i  N0 on day
We denote the maximum elapsed driving time without break
by F and the duration of a break by G .
t  Ti ,
t
Let K denote the set of vehicles. For each vehicle k  K ,
[ gl , hlt ] The start time and the latest ending time of
let Qk and Pk denote the capacity in weight and in volume, driver l  D on day t  T ,
IJTRD | Mar-Apr 2016
Available [email protected] 251
 International Journal of Trend in Research and Development, Volume 3(2), ISSN: 2394-9333
www.ijtrd.com
it
Delivery quantity for customer i on day t  Ti ,   qit xijkt  Qk  k  K, t  T
iN jN0
(7)
H The maximum working duration for each driver

F
over the planning horizon,
The maximum elapsed driving time without
 px
iN jN0
t t
i ijk  Pk  k  K, t  T (8)
break,
G The duration of the break for drivers, u tjk  uikt  cij  M (1  xijk
t
)  Mzikt
A1, A2, A3 Costs i, j  N0, k  K, t  T (9)
u  cij  M (1  x ) i, j  N, k  K, t  T
t
jk
t
ijk (10)
Variables:
Binary
uikt  c x
jN0
t
ij ijk  F  Mzikt i  N0, k  K, t  T(11)
t
xilk Equal to 1 if vehicle k  K travels from node
v  vikt  dit  cij  G  zikt  M (1  xijk
t
jk
t
)
i  N0 to j  N0 with driver l  D on day i, j  N0, k  K, t  T (12)
t T , bi  v  ai t
i  N, k  K, t  Ti (13)
ik
wkt Equal to 1 if vehicle k  K delivers order,
v0t k   ( glt  ylkt ) k  K, t  T (14)
z t
ik Equal to 1 if vehicle k  K takes break after lD

serving node i  N0 on day t T , vnt 1,k   (hlt  yikt ) k  K, t  T (15)

lD
ylkt Equal to 1 if vehicle k  K is assigned to rlt  vnt 1,k  glt  M (1  ylkt )
driver l  D on day t  T
 l  D, k  K, t  Ti (16)
Continuous r
tTl
l
t
H lD (17)
v t
ik The time at which vehicle k  K starts service t
xilk , wit , zikt , ylkt {0,1}
at node i  N0 on day t T ,
i, j  N0, l  D, k  K, t  T (18)
uikt The elapsed driving time of vehicle k  K at v , u , rl  0 i, j  N0, l  D, k  K, t  T (19)
t
ik
t
ik
t

node i  N0 after the previous break on day The objective function (1) minimizes the total cost
incurred over the planning horizon.
t T ,
Constraints (2) to ensure that each customer must be
rl t
The total working duration of driver l  D on visited by one vehicle on each of its delivery days. Constraints
day t T . (3) impose that each customer node is visited at least once.
Flow conservation of vehicles is presented in Constraints (4).
The variables x and w need to be linked as shown in
III. THE MODEL Constraints (5). While Constraints (6) is to make sure that
every vehicle delivers at least one order per customer.
The mathematical formulation for this problem can be Constraints (7-8) guarantee that the vehicle capacities are
formulated as a mixed integer programming model. respected in bothe weight and volume.
Constraints (9-10) define the elapsed driving time as a
function of binary variable zikt Meaning that, for the vehicle
k travelling from customer node i to j on day t, the elapsed
driving time at j equals the elapsed driving time at i plus the
driving time from i to j (i.e., u tjk ≥ uikt + cij ) if the vehicle
(1)
does not take a break at customer i (i.e., zikt = 0). Otherwise, if
Subject to : the vehicle takes a break at customer i (i.e., zikt = 1), the
 xojkt  1
jN
 k  K, t  T (2) elapsed driving time at j will be constrained by (10) which
make sure it is greater than or equal to the travel time between
x
kK jN0
t
ijk 1  i  N, t  Ti (3) i and j (i.e., u tjk ≥ cij). Constraints (11) guarantee that the
elapsed driving time never exceeds an upper limit F by
 xijkt   xtjik  0 i, j  N0, t  T
kK kK
(4) imposing a break at customer i (i.e., zikt = 1) if driving from

x  w
customer i to its successor results in a elapsed driving time
t
ijk
t
k i, j  N0, t  T (5)
kK kK Constraints (12) determine the time to start the service at

 w 1
kK
t
k tT (6)
each customer. If node j is visited immediately after node i,
the time v tjk to start the service at j should be greater than or

IJTRD | Mar-Apr 2016

Available [email protected] 252
 International Journal of Trend in Research and Development, Volume 3(2), ISSN: 2394-9333
www.ijtrd.com
equal to the service starting time vik at i plus its service model. The heuristic method, called neighborhood search, can
t
described as follows.
duration d it , the extra service time e  pit if i is visited by an Stage 1.
Step 1. Get row i* the smallest integer infeasibility, such that
inappropriate vehicle (i.e., wtj = 1), the travel time between  i*  min{ fi ,1  fi }
the two customers cij , and the break time G if the driver (This choice is motivated by the desire for minimal
t deterioration in the objective function, and clearly
takes a break after serving I (i.e., z ik = 1). Constraints (13) corresponds to the integer basic with smallest integer
make sure the services start within the customers’ time infeasibility).
window. Step 2. Do a pricing operation
1
vi*  ei* B
T T
Constraints (14-15) ensure that the starting time and
ending time of each route must lie between the start working This Step is to find the direction vector.
time and latest ending time of the assigned driver. Constraints
Step 3. Calculate  ij  vi* j
T
(16) define the working duration for each driver on every
workday, which equals the time the driver returns to the depot With corresponds to
minus the time he/she starts work. Constraints (17) make sure  d j 
that the drivers work for no more than a maximum weekly min  
working duration. Constraints (18-19) define the binary and
  ij 
j

nonnegative variables used in this formulation.

Calculate the maximum movement of nonbasic j at
IV. NEIGHBORHOOD SEARCH lower bound and upper bound.
An optimal solution of a linear programming problem can Otherwise go to next non-integer nonbasic or
be detected using the reduced gradient vector. However, superbasic j (if available). Eventually the column j*
generally, in integer programming the reduced gradient is to be increased form LB or decreased from UB. If
vector, is not available, even though the problems are convex. none go to next i*.
Thus we need to impose a certain condition for the local Step 4.
testing search procedure in order to assure that we have Solve Bj* = j* for j*
obtained the “best” suboptimal integer feasible solution. Step 5. Do ratio test for the basic variables in order to stay
feasible due to the releasing of nonbasic j* from its
[1] has proposed a quantity test to replace the pricing test bounds.
for optimality in the integer programming problem. The test is Step 6. Exchange basis
conducted by a search through the neighbors of a proposed Step 7. If row i* = {} go to Stage 2, otherwise
feasible point to see whether a nearby point is also feasible Repeat from step 1.
and yields an improvement to the objective function.
Let []k be an integer point belongs to a finite set of Stage 2. Pass1 : adjust integer infeasible superbasics (if any)
neighborhood N([]k) We define a neighborhood system by fractional steps to reach complete integer
associated with []k, that is, if such an integer point satisfies feasibility.
the following two requirements Pass2 : adjust integer feasible superbasics. The
1. if []j  N([]k) then []k  []j, j  k. objective of this phase is to conduct a highly
2. N([]k) = []k + N(0) localized neighborhood search to verify local
optimality.
With respect to the neighborhood system mentioned
above, the proposed integerizing strategy can be described as CONCLUSIONS
follows. This paper was intended to develop efficient technique for
Given a non-integer component, xk, of an optimal vector, solving one of the most economic importance problems in
xB. The adjacent points of xk, being considered are [xk] dan [xk] optimizing logistic systems. The aim of this paper was to
+ 1. If one of these points satisfies the constraints and yields a develop a model of split delivery vehicle routing with Time
minimum deterioration of the optimal objective value we Windows, Fleet and Driver Scheduling Problem This problem
move to another component, if not we have integer-feasible has additional constraint which is the limitation in the weight
solution. and volume of vehicles. The proposed algorithm employs
nearest neighbor heuristic algorithm for solving the model.
Let [xk] be the integer feasible point which satisfies the
above conditions. We could then say if [xk] + 1 N([xk]) References
implies that the point [xk] + 1 is either infeasible or yields an [1] H. E Scarf., Testing for optimality in the absence of convexity in
inferior value to the objective function obtained with respect Walter P Heller, Ross M Starr and David A Starrett (Eds) (Cambridge
University Press), 1986, pp. 117-134, 1986.
to [xk]. In this case [xk] is said to be an “optimal” integer [2] J.-F. Cordeau, G. Laporte, M.W.F. Savelsbergh, and D. Vigo. Vehivle
feasible solution to the integer programming problem. Routing. In Barnhart, C. and Laporte, G., editors, Transportation,
Obviously, in our case, a neighborhood search is conducted handbooks in Operation Research and Management Science, Pages
through proposed feasible points such that the integer feasible 367-428. North-Holland, Amsterdam, 2007.
solution would be at the least distance from the optimal [3] O. Br ysy and M. Gendreau. “Vehicle Routing Problem with Times
Windows, Part II: Metaheuristics”. Transportation Science, 39(1),
continuous solution. 2005, pp. 104-118.
[4] P. Toth and D. Vigo. “The Vehicle Routing Problem Society for
V. THE ALGORITHM Industrial and Applied Matheamtics”, Philadelpiha, PA, 2002.
[5] G.B. Dantzig and J.H. Ramser, “The truck dispatching problem”,
This is a combination of exact and heuristic method. The
Management Science, Vol. 6, 1959, pp. 80.
exact method is used for solving the relaxed problem of the

IJTRD | Mar-Apr 2016

Available [email protected] 253
 International Journal of Trend in Research and Development, Volume 3(2), ISSN: 2394-9333
www.ijtrd.com
[6] Gendreau, M., Tarantilis, C.D., “Solving large-Scale Vehicle Routing [22] M. E. McNabb, J. D. Weir, R. R. Hill, and S. N. Hall., “Testing local
Problems with Time Windows: The State-of-the-Art”. CIRRELT-2010- search move operators on the vehicle routing problem with split
04. Universite´ de Montreal, Canada, 2010. deliveries and time windows”, Computers & Operations Research 56,
[7] Toth, P., Vigo, D. (2nd Eds.). “Vehicle routing : Problems, Methods, and 2015, 93-109.
Applications, Society for Industrial and Applied Mathematics & [23] C. Archetti, N. Bianchessi, and M.G. Speranza. “Branch-and-cut
Mathematical Optimization Society”, Philadelphia PA, 2014. algorithms for the split delivery vehicle routing problem”. European
[8] M. Dror and P. Trudeau, “Savings by Split Delivery Routing,” Journal of Operational Research, 2014.
Transportation Science, Vol. 23, No. 2, 1989, pp. 141-145. [24] C. Archetti, N. Bianchessi, and M.G. Speranza., “A branch-price-and-
[9] M. Dror, G. Laporte and P. Trudeau, “Vehicle Routing with Split cut algorithm for the commodity constrained split delivery routing
Deliveries,” Discrete Applied Mathematics, Vol. 50, No. 3, 1994, pp. problem”, Computers & Operations Research, 54, 2015, pp. 1-10.
239-254. [25] L. Berbotto, S. Garcia, and F.J. Nogales., “A randomized granular tabu
[10] C. Archetti and M. G. Speranza, “The Split Delivery Ve-hicle Routing search heuristic for the split delivery vehicle routing problem”. Annals
Problem: A Survey. The Vehicle Routing Problem: Latest Advances and of Operations Research, 2013, pp. 1-21.
New Challenges,” Operations Research/Computer Science Interfaces [26] M. Boudia, C. Prins, and M. Reghioui., “An effective memetic
Series, Vol. 43, Part I, 2008, pp. 103-122. algorithm with population management for the split delivery vehicle
[11] C. Archetti, M. Savelsbergh and A. Hertz, “A Tabu Search Algorithm routing problem”. In Hybrid Metaheuristics, pages 16-30. Springer,
for the Split Delivery Vehicle Routing Problem,” Transportation 2007.
Science, Vol. 40, No. 1, 2006, pp. 64-73. [27] A. Ceselli, G. Righini, and M. Salani., “A column generation algorithm
[12] C. Archetti, M. Savelsbergh and M. G. Speranza, “To Split or Not to for a rich vehicle-routing problem”. Transportation Science, 43(1),
Split: That Is the Question,” Transportation Research Part E, Vol. 44, 2009, pp. 56-69.
No. 1, 2008, pp. 114-123. [28] G. Desaulniers., “Branch-and-price-and-cut for the split-delivery vehicle
[13] C. Lee, M. A. Epelman, C. C. White and Y. A. Bozer, “A Shortest Path routing problem with time windows”. Operations Research, 58(1),
Approach to the Multiple-Vehicle Routing Problem with Split Pick- 2010, pp. 179-192.
Ups,” Transportation Research Part B, Vol. 40, No. 4, 2006, pp. 265- [29] M. Dror and L. Levy., “A vehicle routing improvement algorithm
284. comparison of a greedy and a matching implementation for inventory
[14] J. M. Belenguer, M. C. Martinez and E. A. Mota, “Lower Bound for the routing”. Computers & Operations Research, 13(1), 1986, pp. 33-45.
Split Delivery VRP,” Operations Research, Vol. 48, No. 5, 2000, pp. [30] M. Jin, K. Liu, and B. Eksioglu., “A column generation approach for the
801-810. split delivery vehicle routing problem”. Operations Research Letters,
[15] M. Jin, K. Liu and R. O. Bowden, “A Two-Stage Algo-rithm with Valid 36(2), 2008, pp. 265-270.
Inequalities for the Split Delivery Vehi-cle Routing Problem,” [31] E. Mota, V. Campos, and _A. Corberan., “A new metaheuristic for the
International Journal of Production Economics, Vol. 105, No. 1, 2007, vehicle routing problem with split demands”. In Evolutionary
pp. 228-242. Computation in Combinatorial Optimization, pages 121-129. Springer,
[16] S. Chen, B. Golden and E. Wasil, “The Split Delivery Vehicle Routing 2007.
Problem: Applications, Test Problems, and Computational Results,” [32] G. Sierksma and G.A. Tijssen., “Routing helicopters for crew exchanges
Networks, Vol. 94, No. 4, 2007, pp. 318-329. on off-shore locations”. Annals of Operations Research, 76, 1998, pp.
[17] R.E. Aleman and R.R. Hill., “A tabu search with vocabulary building 261-286.
approach for the vehicle routing problem with split demands”. [33] J.H. Wilck-IV and T.M. Cavalier., “A construction heuristic for the split
International Journal of Metaheuristics, 1(1), 2010, pp. 55-80. delivery vehicle routing problem”. American Journal of Operations
[18] R.E. Aleman, X. Zhang, and R.R. Hill., “An adaptive memory algorithm Research, 2, 2012, pp. 153.
for the split delivery vehicle routing problem”. Journal of Heuristics, [34] Gendreau, M., P. Dejax, D. Feillet and C. Gueguen, “Vehicle routing
16(3), 2010, pp. 441-473. with time windows and split deliveries, Technical Report, Technical
[19] C. Archetti and M.G. Speranza., “Vehicle routing problems with split report 2006-851, Laboratoire d’Informatique d’Avignon, 2006.
deliveries”. International Transactions in Operational Research, 19(1- [35] A. F. Han, and Y. Chu., “A multi-start heuristic approach for the split
2), 2012, pp. 3-22. delivery vehicle routing problem with minimum delivery amounts”,
[20] C. Archetti, M.G. Speranza, and A. Hertz., “A tabu search algorithm for Transportation Research Part E 88, 2016, pp. 11-31.
the split delivery vehicle routing problem”. Transportation Science, [36] Vidal, T., T. G. Crainic, M. Gendreau, C. Prins. “Heuristics for multi-
40(1), 2006, pp. 64-73. attribute vehicle routing problems: A survey and synthesis.” CIRRELT-
[21] M. Salani and I. Vacca C. , “Branch and price for the vehicle routing 2012-05.
problem with discrete split deliveries and time windows”. Report [37] Kumar, S. N., R. Pannerselvam, “A survey on the vehicle routing
TRANSP-OR 091224, Ecole Polytechnique, 2009. problem and its variants”, Intelligent Information Management, 4,
2012,pp. 66-74..

IJTRD | Mar-Apr 2016

Available [email protected] 254