You are on page 1of 9

EUROPEAN

JOURNAL
OF OPERATIONAL
RESEARCH
ELSEVIER European Journal of Operational Research 97 (1997) 571-579

Theory and Methodology

A branch and bound algorithm for the traveling purchaser


problem
Kashi N. Singh a, *, Dirk L. van Oudheusden b
a Department of Mechanical Engineering, Bihar College of Engineering, Patna 800 005, India
b Division of Industrial Management, Catholic University ofLouvain, B 3001 Leuven, Belgium

Abstract

An important generalization of the traveling salesman problem called the traveling purchaser problem is considered. A
branch and bound algorithm which solves a related simple plant location problem for calculating the bounds is proposed for
this problem. Computational experiments with this algorithm provide the evidence that moderate size problems of upto 25
cities and 100 commodities can be solved in reasonable computation time. The problem is shown to have applications in the
areas of scheduling, warehousing and routing. © 1997 Elsevier Science B.V.

Keywords: Traveling purchaser problem; Traveling salesman problem; Simple plant location problem; Heuristic; Algorithm

1. Introduction price. Also, we assume that triangular inequality


holds in travel costs.
W e consider a problem in which a set 1 = The traveling purchaser problem (TPP) as de-
{1, 2 . . . . . m} of m cities and a set J = {1, 2 . . . . . n} scribed above is NP-hard as one of its special cases
of n commodities are given. Also, we are given c u, where m = n and each city carries only one com-
the cost of commodity j at city i, and hi~, the cost modity reduces to a traveling salesman problem
of travel from city i to city k. A purchaser starts (TSP).
from his home city (say city 1), travels to a subset of The name, traveling purchaser problem, was
m cities and purchases each of the n commodities, coined by Ramesh (1981) who presented a lexico-
and finally returns to his home city. The problem is graphic search procedure to solve the problem ex-
to find a tour for the purchaser such that the sum of actly. He could solve problems of only upto 12 cities
the travel and purchase costs is minimized. and 10 commodities. Prior to this Burstail (1966)
To avoid explicit consideration of trivial cases, encountered a problem o f similar structure in a tube
we assume that each commodity is available in at manufacturing firm. He developed a heuristic which
least one of the m cities, and that none of the cities was further commented on by Lomnicki (1966).
can provide each o f n commodities at the cheapest Buzacott and Dutta (1971) developed an exact proce-
dure based on dynamic programming for a job se-
quencing problem having similar structure.
* Corresponding author. Present address: National Institute of
Industrial Engineering (NITIE), Vihar Lake, Mumbai 400087, Golden et al. (1981) developed a heuristic ap-
India. proach to solve the TPP. Their heuristic was, later

0377-2217/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved


PII S0377-221 7(96)003 13-X
572 K.N. Singh, D.L. van Oudheusden/ European Journal of OperationalResearch 97 (1997) 571-579

on, modified by Ong (1982) but in the absence of Yik ~ { 0 , 1} i = 1 . . . . . m;


any powerful exact procedure it was not known how
good the solutions given by these heuristics would k = 1. . . . . m (5)
be. and cities finally selected must be
In the present paper a branch and bound algorithm connected by a single tour (6)
for the TPP is developed. The basic idea is that the
The objective function (1) represents the sum of
two aspects of the problem, viz., selection of a
commodity cost and travel cost. The constraints (2)
subset of m cities for the tour and determination of
express that each commodity has to be purchased at
an optimal tour of these cities, are successfully em-
a single city. The constraints (3) ensure that only the
bedded into one. While for the resolution of the
cities in the optimal tour are considered for purchas-
former, similarity of the TPP with the simple plant
ing commodities. Constraints (4) provide that a com-
location problem (SPLP) is exploited, for the latter, a
modity is either purchased at a given city or not.
branching rule similar to that of Little et al. (1963)
Similarly, constraints (5) indicate that in an optimal
for the TSP is used. The algorithm is found to have
tour a city either immediately follows a given city or
resolved comparatively large size problems (with 25
not.
cities and 100 commodities) on an IBM 3031 com-
This is to be noted that the above problem, (1) to
puter in a reasonable computation time.
(6), represents a more general formulation of the
TPP than that described in the beginning as none of
the cities (such as home city) is constrained to be in
2. M a t h e m a t i c a l formulation
the tour. It will be seen later that this can be ensured
The constituents of the problem are without any extra difficulty.
m : the number of cities indexed by i, i ~ l =
{1 . . . . . m}.
n : the number of commodities indexed by j, j ~ J 3. A branch and bound algorithm
= {1 . . . . . n}.
cij : cost of commodity j at city i. The TPP as formulated above is difficult to solve
hik : cost of travel from city i to city k, k ~ I. by a direct approach. Hence, a branch and bound
Let us introduce the Boolean variable: algorithm is devised. The basic idea is to break up
Xi~ equal to 1 if commodity j is purchased from city the set of all possible tours into smaller and smaller
i and equal to 0 otherwise. subsets and to calculate for each of them a lower
Yik equal to 1 if city k immediately follows city i in bound on the sum of the tour travel cost and the
the optimal tour and equal to 0 otherwise. commodity cost. The lower bound is calculated by
Now the TPP can be represented as follows: solving a relaxation of the problem which is similar
to an SPLP. The bounds guide the partitioning of the
Minimize ~,~ ~ cijXij "[- ~ ~ hikYik (l) subsets and eventually identify an optimal solution
i=1 j = l i=1 k = l when a subset is found that contains a single tour
m and whose bound is less than or equal to the lower
Subject to ~ Xij -~- 1 j = 1 . . . . . n (2) bounds for all other subsets.
i=l

k~l
r,,! i = 1 . . . . . m;
3.1 Lower bounds

m j = 1 ..... n (3) To calculate lower bounds some of the con-


straints, (2) to (6), are relaxed and the resulting
k=l
problems are solved. Let us define variables
X i j ~ {0, 1} i = 1 . . . . . m;
j- 1 . . . . . n (4) Yi = ~ Y i k = ~ Y k i (7)
k=l k=l
K.N. Singh, D.L. van Oudheusden/ EuropeanJournal of OperationalResearch 97 (1997)571-579 573

such that (7) ensure that in the optimal tour a city i / all "~
must have a single immediate predecessor city and a tours ]
single immediate successor city. Now constraints
(3)-(5) imply

Xij ~ Yi (8) [ i, k
\
and
/
{o, 1} (9) /

The Boolean variable, Yi, will be equal to 1 if city i (~/) t,p !


is included in the optimal tour and equal to zero
otherwise. Further, let us define
Fig. I. Branching of a tree.
fi = rain { hik } (10)
k

fi, here, represents the minimum cost to be incurred


for traveling out of city i. Then it is obvious that
1) represents all tours which include the city pair
(i, k), whereas the node containing i, k represents
i=1 k = l i=1 all tours that do not include city pair (i, k). If a
further branching is done at the i, k node, the node
Hence, the right hand side of (11) will give a lower
containing t, p represents all tours that include (i, k)
bound on the total travel cost. Now the solution to
but not (t, p), whereas t, p represents all tours that
the following problem will give a lower bound on
include both (i, k) and (t, p).
the minimum objective function value of the TPP
By tracing from a given node, P, back to start,
represented by (1)-(6).
we can easily know which city pairs are committed
to appear in the tours of P and which are forbidden
~ cijXij-~- ~ f i Y i (12) to appear. If the branching is carried further, some
i=1 j = l i=1
node will eventually represent a single tour of com-
Subject to (2), (4), (8) and (9) mitted city pairs. So the main decision involved at
each branching step is how to choose a city pair
(t, p) to base the branching on. In the proposed
The above problem has the same form as the
algorithm, at each step the city pair (t, p) is chosen
SPLP-model of Efroymson and Ray (1966) and can
in such a way that the problem with this pair ex-
be easily solved by any of the available approaches
cluded or t,(t(t(~) will yield as large a bound as
(e.g., Bilde and Krarup, 1977; Erlenkotter, 1978). In
possible. This branching decision rule is similar to
fact one of the special cases of TPP, wherein hik = C
that of Little et al. for the TSP.
(a constant), i ~ I, k ~ I, results in an SPLP.
When a node, P, branches into two further nodes,
the node with the newly committed city pair will be
3.2 Branching called Q and the node with the newly forbidden city
pair C9.
The initial solution to the above problem consti-
tutes the root of the search tree. This is also called
the initial node of the tree and it represents the set of
3.3 Flow chart of the algorithm
all tours. The process of partitioning for this set of
all tours into disjoint subsets is represented as
branching of the tree and the subsets of tours as Each individual step of the algorithm has been put
nodes of the tree. The node containing i, k (see Fig. in a separate box in the flow chart (see Fig. 2) and
574 K.N. Singh, D.L. van Oudheusden / European Journal of Operational Research 97 (1997) 571-579

' 'START )

, Initialize
l
If = Travel Cost Matrix
C = Commodity Cost Matrix
P = 1 (all tours), 7,0 = ~

2 1
Calculate fi ~ min { hlk I-
Solve SPLP k
Label P by w(P) = Objective
function value of SPLP
Level = 1

Yes Yes Z0 = w (P)


!
Save T o u r
Solution is optimal

.~o
()
~3

[
Choose city-pair (t, p) for tree extension
such that node with this Pair excluded or Solution is I
( I ~ will yield as large a bound as Z0 = w (O)
optimal
possible. Save tour

5" 1
Make a branch from P t o ~ 7 the CP node.
ft = f't, the next higher cost. Solve SPLP.
10
Set up H for P :
No

1. H = OriginalTravel Cost Matrix


L a b e l ~ by w ( ( ~ - = objective function 2. Read pairs (i, k) committed to be in tours of P.
value of SPLP. Restore original value of ft. 3. For each such (i, k) delete row i and column k
of H. Prevent subtours if necessary. For each
t, p prohibited from tours of P, set htp = m.
4. Find values of fi.
5. Label this node byw (P) = w (O)~

1
Make a branch from P to Q, the t, p node.
Delete row t and column p in H. Make
necessary changes in fi values. Prevent
subtour of connected path if required.
Solve SPLP and label
Q by w (O) = Objective function value of
SPLP.
Level = Level + 1

()
Yes

Yes
No

F i g . 2. F l o w c h a r t o f t h e a l g o r i t h m .
K.N. Singh, D.L. van Oudheusden / European Journal of Operational Research 97 (1997) 571-579 575

the step number is marked thereon. The working of Table 1


the algorithm is explained through an example prob- Computational results for problems with asymmetric travel costs.

lem given in the Appendix. Problem Problem size t~ S t2


In a nutshell the following are the main features No. m n
of the proposed branch and bound algorithm, l 10 30 0.137 0.105 -
a) At each node the lower bound is determined by 2 10 50 1.095 0.746 0.144
solving a related SPLP and the technique used is the 3 10 100 3.126 0.401 0.393
dual-based method of Erlenkotter (1978). The method 4 15 10 1.709 0.662 1.092
5 15 50 14.797 1.235 2.035
constructs optimal or near optimal solutions in 6 15 100 51.987 2.033 0.582
O(nm 2) instructions (additions, subtractions and 7 20 10 1.817 0.025 0.113
comparisons). Since we are interested only in good 8 20 30 18.537 6.695 6.815
bounds we do not solve SPLP to optimality at each 9 20 50 36.518 2.944 6.447
step and hence the branch and bound phase of this to 20 100 124.593 7.162 2.732
11 25 10 6.655 5.685 0.592
method is not used for solving SPLP. We could see, 12 25 30 63.636 17.345 22.199
however, that the dual-ascent phase itself produced 13 25 50 85.980 1.600 12.020
optimal solutions in most of the cases.
tt: mean computation time (sec) excluding 1 / 0 times to obtain an
b) For branching, a city pair is selected such that optimal solution.
the problem with this pair excluded will yield as t2: mean computation time (sec) to obtain the first feasible
large a bound as possible. solution.
c) Fathoming of a node is done either by bound- s: standard deviation of computation times.

ing i.e., when the bound exceeds the cost of the best
solution obtained so far or on getting a feasible
solution. For a detailed discussion of each aspect of where a is an integer whose value does not influence
the algorithm the reader may refer to Singh (1985). the optimal solution as all the commodities are to be
purchased. The proportion of commodity costs dif-
ferent from M is around 50%. A problem size is
4. Computational results specified by the number of cities, m, and the number
of commodities, n. A total of 65 problems (5 prob-
A Computer program in FORTRAN, called TRA- lems of each size) with asymmetric travel costs, and
PUR, of the above algorithm is written and tested on of 10 to 25 cities and 10 to 100 commodities are
an IBM 3031 computer. Since standard test problems generated and solved. As the data on computation
are not available, most of the cost data are randomly times for any other approach are not available, nu-
generated. As some of the applications of TPP can merical comparison was not possible.
be found in job sequencing on machines with se- Table 1 presents the mean computation time (t~)
quence-dependent changeover costs and since for each size of problem solved by TRAPUR. The
changeover costs need not be symmetric, we experi- table also includes the standard deviation (s) of
ment with asymmetric travel costs first. Later on, we computation times, and the mean CPU time Oz) for
experiment with symmetric travel costs as well. The obtaining the first feasible solution. The first feasible
results of these experiments are summarized in Table solution values were always within 5% and in many
1 and Table 2. cases within 1% of the respective optimal solution
Asymmetric travel costs for problems of various values. The results indicate that 10-city problems
size are generated in the same manner as has been with upto 100 commodities could be solved in a few
done for a similar scheduling problem in Hansen and seconds. Similarly, problems with 15 to 25 cities and
van Oudheusden (1972). Travel costs are integers 10 commodities could be solved in small computa-
randomly generated from a uniform distribution in tion times. Computation times start increasing when
the range (15, 30). Commodity costs different from problems of size (15 × 50) or bigger are solved. For
arbitrary large value, M, are also integers generated problems with same number of commodities, addi-
from a uniform distribution in the range (a, a + 10), tion of 5 cities increases the computation time by
576 K.N. Singh, D.L. van Oudheusden / European Journal of Operational Research 97 (1997) 571-579

Table 2
Computational results for problems with symmetric travel costs
Problem No Problem size t I (sec) t 2 (sec) Average relative Average relative
travel cost commodity cost
m n
TRAPUR/GS TRAPUR/GS

1 10 15 0.799 0.064 0.675 1.00


2 10 30 4.017 0.130 0.914 0.989
3 10 50 6.853 0.182 0.904 1.00
4 15 15 5.002 0.201 0.871 1.00
5 15 30 15.943 0.243 0.859 0.992
6 15 50 57.839 0.360 0.812 1.072
7 20 15 16.189 0.347 0.657 1.080
8 20 30 33.961 0.399 0.691 0.961
tl: mean computation time for TRAPUR (IBM 3031).
t2: mean computation time for GS (IBM 3031).

about 2 to 3 times, and for problems with same and 50 commodities. For problems of larger dimen-
number of cities, addition of 50 commodities in- sions, computation times may be comparatively
creases the computation time by about 3 to 4 times. higher.
For generating symmetric problems, hik values
are taken from the travel cost data for the 33-city
TSP given in Karg and Thompson (1964). cij values 5. Applications and extensions
different from M are integers randomly generated
from a uniform distribution in the range (0-500). A The structure of the traveling purchaser problem
total of 40 problems with 10 to 20 cities and 15 to 50 applies to many real-life problems. We discuss here
commodities are solved. These problems are also some of its potential applications to job scheduling,
solved by a generalized savings (GS) heuristic devel- warehousing and routing. In the later portion of this
oped by Golden et al. and the solutions given by the section we describe a possible extension of this
GS heuristic are compared with the optimal solutions problem.
obtained by TRAPUR. Table 2 provides the data on
mean computation time (t l) for TRAPUR, mean 5.1. Job scheduling
computation time (t z) for the GS heuristic and the
average relative travel and purchase costs. As mentioned earlier, the job scheduling problem
The results of Table 2 indicate that solution times considered by Burstall (1966) and later by Buzacott
for problems with symmetric travel costs are gener- and Dutta (1971) has the same structure as the TPP.
ally higher as compared to those for asymmetric cost hik, in this case will represent the changeover cost
problems. The reason may be that for every feasible from state i to state k, and cij, the cost of processing
solution to such a problem there is always another job j when the machine is in state i. The home city
solution in reversed sequence but having the same (city 1) of the TPP may represent a dummy state of
objective value. This results in more number of the machine in which a dummy job, (n + 1), is
alternatives to be examined. As regards the quality of processed at cost zero. The changeover costs from
solutions produced by the heuristic it is seen that this dummy state to other states of the machine may
quality deteriorates with the increase in problem size, be set equal to zero. Then the solution to this prob-
the only exception being the (10 × 15) problems. lem (as obtained by TRAPUR) will give a sequence
The heuristic produced solutions which are 10 to of selected states in which different jobs are to be
35% off the respective optimal solutions. processed.
The overall conclusion that can be drawn from However, in this procedure the initial bound given
these computational results is that the code, TRA- by the SPLP will be loose as all f, values will be
PUR, can efficiently solve TPP with upto 20 cities zero due to the presence of zeros in a row and a
K.N. Singh, D.L. van Oudheusden / E u r o p e a n Journal o f Operational Research 97 (1997) 5 7 1 - 5 7 9 577

column of the {hik} matrix. Subsequently, bounds 5.3. Routing


may improve. If, however, the initial or final state of
the sequence is known, the algorithm can be used as One of the routing applications of the TPP is in
it is and no dummy state will be needed at all. In that the school bus problem. A school bus has to pass by
case it is enough to modify the steps leading to the many points to gather children for the school. Let hik
construction of a tour in such a way that a path is represent the cost of driving from point i to point k.
generated starting or ending at the specified node. Also, let cij be the distance from child j ' s house to
point i, weighted in order to be expressed in the
same unit as h/k. Now the problem is to determine
5.2. Warehousing the tour of the bus which minimizes the sum of the
driving costs, h i k , and the weighted distances, cij,
One of the most fundamental item retrieval prob- walked by the children from their houses to the
lems associated with warehousing is the 'order-pick- nearest points of the tour. Obviously, this problem
ing' problem. An order consists of a subset of items has the same form as the TPP.
stored in a warehouse. On receiving an order, the Current et al. (1984) introduced a problem called
warehouse dispatches a vehicle from the picking area shortest covering path problem (SCPP) which is a
to 'pick' the items in the 'order' and transport them synthesis of the set covering location problem and
back to the shipping area. The objective is to mini- the shortest path problem. The SCPP determines the
mize distance (or time) traveled by the vehicle. shortest path between two given nodes in a graph
For the situation when an item is stored in a such that all the demand nodes are covered. A
single location area of the warehouse, the order-pick- demand node is considered covered if either it is
ing problem has been recognized as a variant of the directly on the shortest covering path or if it is
TSP. However, when an item is stored in more than within a predetermined maximum distance from a
one location area, the problem has the structure of a node on the path. The SCPP can be shown to be a
TPP. Let i = 1, 2 . . . . . m, denote the location areas special case of the TPP. hik, as usual, will represent
of a warehouse with location 1 representing the the distance (cost) from node i to node k and cij
shipping area. Let j = 1, 2 . . . . . n, denote the items may be given by
stored in these areas. Then the two cost matrices for 0 if jth node is covered by ith node
the TPP may be constituted as follows. ciJ= oc otherwise
H = {hik} will be an (m X m) matrix consisting of
traveling distances (costs) from one location area to Hence, both H and C will be (n × n) matrices,
other. C = { c J will be an (m × m) matrix given by where n is the total number of nodes in the graph. It
is interesting to note that the bounding problem
0 if item j is stored in area i reduces to an SCP in this case as well.
Ci j = o¢ otherwise Many potential applications of the SCPP to rout-
ing and network design problems have been dis-
Thus the order-picking problem becomes a special cussed in Current et al. Due to the above relationship
case of TPP for which cli values are either 0 or oc. between SCPP and the TPP all these also apply to
This special feature may give some computational the latter. The design of a subway line or rail line
advantage when the problem is solved by TRAPUR. may be of particular interest. The stations may be
This is so because the bounding problem will reduce located at some of the population centers while the
to a set covering problem (SCP) which is easier to neighboring areas will be covered by these stations.
solve. Moreover, sometimes two or more items share Similarly, the same principle may be applied to
the same location areas in a warehouse. These are develop irrigation network in a given region or a
such items which are needed, most of the times, road network between two principal cities.
simultaneously. In such cases each group of items is One of the two problems suggested by Ramesh
stored in more than one area and the order-picking (1981) as possible extensions of the TPP is the
problem is again an instance of TPP. dynamic TPP. In this case, a time element is intro-
578 K.N. Singh, D.L. van Oudheusden / European Journal o f Operational Research 97 (1997)571-579

duced in the problem. All the cost parameters change network design problems may be topics of further
from one time period to another. This can also be study.
formulated in the same manner as its static counter-
part and the same solution procedure can be applied.
The only change needed is that a dynamic SPLP is to Acknowledgements
be solved for determining the bound. A dynamic
SPLP can be transformed into an equivalent static The work reported in this paper was carried out
SPLP and it may be attempted by the same dual-based while the first author was a doctoral student at the
method of Erlenkotter (1978). However, a slightly Asian Institute of Technology, Bangkok. The finan-
improved procedure which incorporates primal-dual cial support provided for the study by the Govern-
adjustment in the dual ascent phase has been devel- ment of Japan is thankfully acknowledged. The au-
oped by Van Roy and Erlenkotter (1982). Their thors are indebted to the anonymous referees for
D Y N A L O C code based on this algorithm has been their comments which considerably improved the
able to solve (15 × 30) problems of 8 periods in presentation of the paper.
fractions of a second on an IBM 3033 computer. So
this procedure can be embedded to the code, TRA-
PUR, for the solution of a dynamic TPP. Appendix

Each step of the algorithm is now explained


6. C o n c l u d i n g remarks through an example problem taken from Ong (1982).
This is a 5 city-8 commodity problem for which cost
The traveling purchaser problem, an extension of matrices H and C are given in Table 3. A 9th
the well-celebrated traveling salesman problem is commodity (dummy), which is available only at city
considered. Since the problem is NP-hard, possibility 1 at cost zero, is added to ensure that city 1 (home
of a polynomial time algorithm is most unlikely. city) is always in the solution.
Hence, a branch and bound algorithm in which For the example problem f / = mink{h/k} =
bounds are determined by solving a related simple {32, 16, 19, 8, 6}. With f/ as fixed costs and cij as
plant location problem at each node is developed. variable costs, the SPLP solution gives a lower
Computationally, this algorithm is found to have bound of 216 for the TPP. This forms the root of the
considerably extended the size of problems that could tree. The solution indicates that all the 5 facilities are
optimally be solved. The TPP may find applications open, i.e. all the cities are to be visited. This gives a
to many real-life problems, especially in the areas of tour of {(1, 2), (2, 5), (5, 3), (3, 5), (4, 2)} which is
scheduling, warehousing and routing. Two important acyclic and hence the tree will be extended from this
combinatorial problems viz., 'order-picking' problem node having a bound, w ( P ) = 216.
and the shortest covering path problem may be treated We base the branching on a city pair (t, p) such
as the two special cases of TPP. Its extension to the that the tours with city pair (t, p) excluded will give
dynamic case may also be solved by the same proce- as large a bound as possible. For this we calculate
dure. Application of TPP to solve some important the next higher costs, f~', for going out of the cities.

Table 3
(a) Travel cost matrix, H and (b) Commoditycost matrix, C of a 5-city, 8-commodityproblem.
i\k 1 2 3 4 5 i\j 1 2 3 4 5 6 7 8 9
I - 32 54 47 48 1 M M M M M M M M 0
2 18 - 22 25 16 2 M M 17 26 85 M 32 M M
3 39 41 - 28 19 3 42 82 63 94 60 M 12 55 M
4 26 8 30 - 24 4 60 10 57 8 M 95 96 9 M
5 35 17 6 9 - 5 M M M 80 21 16 M 77 M
(a) (b)
K.N. Singh. D.L. van Oudheusden / European Journal of Operational Research 97 (1997) 571-579 579

{(1, 3), (3, 5), (5, 4), (4, 2), (2, 1)} with a cost 243.
Now the left-hand branches are searched till the tree
is complete (see Fig. 3).
An optimal solution to the above problem gives a
2 tour of 1 - 3 - 5 - 4 - 2 - 1 at a travel cost of 108 units.
The total commodity cost is 135 units. Commodity 9
is to be purchased from city 1; commodity 3 from
city 2; commodities 1 and 7 from city 3; commodi-
ties 2, 4, 8 from city 4; and commodities 5, 6 from
city 5.

References

Bilde, O., and Krarup, J. (1977), "Sharp lower bounds and


efficient algorithms for the simple plant location problems",
_ / N_ ~,~ Annals of Discrete Mathematics 1, 79-97.
/'-5 Y-"~ 2,, Burstall, R.M. (1966), " A heuristic method for a job sequencing
problem' ", Operational Research Quarterly 17, 291-304.
2,5 Buzacott, J.A., and Dutta, S.K. (1971), "'Sequencing many jobs
on a multipurpose facility", Naval Research Logistics Quar-
terly 18, 75-82.
An optimal tour is : [(1,3) ~(3~5) ,(5,4] ~(4~2)~ (2~1 ]l Current, J., Revelle, C., and Cohon, J. (1984), "The shortest
covering path problem: an application of locational constraints
Fig. 3. The final tree.
to network design", Journal of Regional Science 24, 161 - 183.
Efroymson, M.A., and Ray, T.L. (1966), " A branch and bound
algorithm for plant location problem', Operations Research 14.
In this case fi' --- {47, 18, 28, 24, 9}. Now t is cho- 361-368.
sen such that Erlenkotter, D. (1978), " A dual-based procedure for uncapaci-
i~A tated facility location", Operations Research 26, 992-1109.
Golden, B., Levy, L., and Dahl, R. (1981), "'Two generalizations
( f ' t - f / ) = E {f/'-f/} of the traveling salesman problem", OMEGA 9, 439-455.
max
Hansen, P., and van Oudheusden, D.L. (1972), "One-machine
where A is the set of facilities (cities) that are open Scheduling with Sequence-dependent change-over costs",
(are to be visited) in the previous solution. P is the Working paper, Free University, Brussels.
city corresponding to which hip = f t . In the given Karg, R.L., and Thompson, G.L. (1964), " A heuristic approach to
solving traveling salesman problem", Management Science
problem (f/' - f / ) = {15, 2, 9, 16, 3} and max(f/' -
10, 225-248.
fi) = ( f 4 - f 4 ) = 16. Hence t = 4. Also, since h42 = Little, J.D.C., Murty, K.G., Sweeney, D.W., and Karel, C. (1963),
f4, P = 2. Hence, the first branching is based on city "'An algorithm for the traveling salesman problem", Opera-
pair (4, 2). tions Research I l, 972-989.
The bound for the node ~ is 232. This is Lomnicki, Z.A, (1966), "Job Scheduling", Operational Research
Quarterly 17, 314-316.
obtained by solving the SPLP after setting ha2 = ac
Ong, H.L. (1982), "Approximate algorithms for traveling pur-
and f4 = 24. The bound for the node (4, 2) is found chaser problem", Operations Research Letters l, 201-205.
to be 231. This is done after deleting the 4th row and Ramesh, T. (1981), "'Traveling purchaser problem", Opsearch
2nd column of the matrix, H. Also, the subtour of 2 18, 78-91.
to 4 is prevented by setting h24 = ~. The values of f/ Singh, K.N. (1985), "The SPLP-model in scheduling and
routing", Doctoral dissertation, IE85-1, Asian Institute of
for this case are {47, 16, 19, 8, 6}.
Technology, Bangkok, Thailand, 55-68.
Now the node (4, 2) is branched on the basis of Van Roy, T.J., and Erlenkotter, D. (1982), " A dual-based proce-
city pair (3, 5). Next we go to the right from 3, 5 on dure for dynamic facility location problem", Management
the basis of (5, 4) where we get a feasible solution, Science 25, 1091-1105.

You might also like