Council for Innovative Research International Journal of Computers & Technology

www.ijctonline.com ISSN: 2277-3061 Volume 3, No. 1, AUG, 2012

Traveling Salesman Problem: A Case Study

Dr. Leena jain Mr. Amit Bhanot
Associate Professor & Head-MCA Global Institute of Management
Global Institute of Management and and Emerging Technologies,
Emerging Technologies, Amritsar, 143 001 (India)
Amritsar, 143 001(India)

ABSTRACT multitude of heuristic algorithms, including genetic

In this paper assignment based integer linear formulation algorithms (GA‟s) [14] for solving TSP [15]. Recently
presented for solving traveling salesman problem. studied approaches to solve TSP includes Knowledge based
Unfortunately, the assignment model can lead to infeasible multiple inversion and knowledge based neighborhood
solutions. Infeasibility removes by introducing additional swapping GA algorithms [16], hybridized GA for integrated
constraints. Then this linear problem solved by open source TSP and quadratic assignment problem [17], two-level GA for
software. clustered TSP and large scale TSP [18], parallel GA program
implementation on multicomputer cluster [19].
General Terms
Optimization technique, Linear Programming, Assignment 3. ASSIGNMENT BASED
problem. TSP, combinatorial optimization problem FORMULATION
Starting from his home, a salesman wishes to visit each of (n
−1) other cities and return home at minimal cost. He must
Keywords visit each city exactly once and it costs cij to travel from city i
Optimization technique, Linear Programming, Assignment to city j.
problem. TSP, combinatorial optimization problem
We may be tempted to formulate his problem as the
1. INTRODUCTION assignment problem:
The Traveling Salesman Problem (TSP) is a classical
combinatorial optimization problem, which is simple to state
but very difficult to solve. The problem is to find the shortest
possible tour through a set of N vertices so that each vertex is xij = 1 if he goes from city i to city j,
visited exactly once. This problem is known to be NP-
complete, and cannot be solved exactly in polynomial time 0 otherwise,
because the number of possible routes increases factorally
with the number of cities. In other words, for 5 cities there are
5! (120) possible routes and for 6 cities there are 6! (720). A
salesman who had to visit the 15 European capitals in this n n
example would have 1,307,674,368,000 (over 1.3 trillion) Minimize cij xij
possible routes to consider. This traditional approach become i 1 j 1
impractical in terms of computer memory and speed
constraints. So to solve this problem different researchers use Subject to:
heuristic, metaheuristic and optimal method like dynamic n
programming, linear programming. In this paper traveling xij 1 ( j = 1, 2, . . . , n),
salesman problem solved like assignment problem using i 1
linear programming approach. The constraints require that
n
the salesman must enter and leave each city exactly once.
Unfortunately, the assignment model can lead to infeasible xij 1 (i = 1, 2, . . . , n),
solutions. Consequently, additional constraints must be j 1
included in order to eliminate subtour solutions. There are a
number of ways to accomplish this. Additional constraints
xi j 0 (i = 1, 2, . . . , n; j = 1, 2, . . . , n).
2. REVIEW OF LITERATURE
The Traveling Salesman Problem (TSP) is one of the The constraints require that the salesman must enter and
important subjects which have been widely addressed leave each city exactly once. Unfortunately, the assignment
extensively by mathematicians and computer scientists [1]. model can lead to infeasible solutions. Infeasibility removes
Many approaches have been tried to solve TSP problem. by additional constraints discussed in next section.
Among them, there are polynomial-sized linear programming
formulation [2], Genetic Algorithm [3,4,5,6], a mixed integer
linear programming formulation and dynamic programming
4. EXPERIMENT SETUP
[7,8,9], NP–complete problem[10,11], Neural Network [12]. To explain the procedure to solve traveling salesman problem
The multiple traveling salesman problem (mTSP) is a we consider the following example in which we have five
cities 1, 2, 3, 4 and 5.
generalization of the well-known traveling salesman problem
(TSP), where more than one salesman can be used in the
solution [13].Historically, researchers have suggested a From 1 2 3 4 5
1 - 3 6 2 3

167 | P a g e www.cirworld.com
Council for Innovative Research International Journal of Computers & Technology
www.ijctonline.com ISSN: 2277-3061 Volume 3, No. 1, AUG, 2012

2 3 - 5 2 3
3 6 5 - 6 4
4 2 2 6 - 6
5 3 3 4 6 -

In the above problem we have 5 cities, let us designate the

visit of ith city to jth city by the decision variable xij,
i=1,2,3,4,5 and j=1,2,3,4,5 but i≠j.

Mathematical Formulation

MIN Cost=
3*x12+6*x13+2*x14+3*x15+3*x21+5*x23+2*x24+3*x25+6*x31+
6*x32+6*x34+4*x35+2*x41+2*x42+6*x43+6*x45+3*x51+3*x52+
4*x53+6*x54;

Subject to Constraints: Fig 2: Output Generated Using Lingo 11 Software

(x12+x13+x14+x15)=1
(x21+x23+x24+x25)=1
(x31+x32+x34+x35)=1
(x41+x42+x43+x45)=1
1 2 3
(x51+x52+x53+x54)=1 1
(x21+x31+x41+x51)=1
(x12+x32+x42+x52)=1
(x13+x23+x43+x53)=1
(x14+x24+x34+x54)=1
(x15+x25+x35+x45)=1 4 5
xi j 0 (i = 1, 2,. .., 5; j = 1, 2, . . . , 5).
Fig 3: disjoint subtours
To solve the ILPP, Software LINGO Version 11 was used.
Figure 1 and Figure 2 illustrate the snap shots of construction Unfortunately, the assignment model can lead to infeasible
of ILPP and generated solution by software LINGO 11, solutions. Infeasibility removes by additional constraints like
respectively. x13+x15+x23+x25+x43+x45>=1;………..(1)

Solution: This inequality ensures that at least one leg of the tour
connects cities 1, 2, and 4 with cities 3 and 5. In general, if a
x12=x24=x41=x35=x53=1 constraint of this form is included for each way in which the
x13=x14=x15=x21=x23=x25=x31=x32=x34=x42=x43= cities can be divided into two groups, then subtours will be
x45=x51=x52=x54=0 eliminated. Now again above ILPP solved with additional
constraint (1) using Software LINGO Version 11 that produce
It is clear from the produced result assignment solution to 1 tour cycle (fig 4).
route the salesman through disjoint subtours of the cities(1-2-
4-1 and 3-5-3) instead of on a single trip or tour (figure 3) Solution:
Minimum Cost z=16
x14=x23=x35=x42=x51=1
x12=x13=x15=x21=x24=x25=x31=x32=x34=x41=
x43=x45=x52=x53=x54=0
Above result produce 1 tour cycle 1-4-2-3-5-1.

Fig 1: Formulation (Lingo 11 Software) 3

Fig 4: One Tour Cycle

168 | P a g e www.cirworld.com
Council for Innovative Research International Journal of Computers & Technology
www.ijctonline.com ISSN: 2277-3061 Volume 3, No. 1, AUG, 2012

[9] R.J.V. Wiel, N.V. Sahinidis, Heuristic bounds and test

5. CONCLUSION problem generation for the time-dependent traveling
In this paper assignment based integer linear formulation salesman problem, J. Trans. Sci., Vol. 29, pp. 167–183,
presented for solving traveling salesman problem. 1995.
Unfortunately, the assignment model can lead to infeasible [10] T. H. Cormen, C. E. Leiserson, R. L. Rivest, and C.
solutions. Infeasibility remove by introducing additional Stein, Introduction to algorithms, (MIT Press, 2001).
constraints. Then this linear problem solved by open source [11] N. T. Shaked, S. Messika, S. Dolev, and J. Rosen,
software. Due to the constraint of length of the paper only one “Optical solution for bounded NP-complete problems,”
case have been reported here. However, more than ten Appl. Opt. 46, pp. 711-724, 2007.
problems have been solved using this approach. [12] S. Kedar, Naphade & Dilek Tuzun, “Initializing the
Hopfield-Tank Network for the TSP using a convex
6. ACKNOWLEDMENT hull: A Computational Study. Proceedings of the
The authors are grateful to Mark Willey Lingo Systems, Inc., Artificial Neural Networks in Engineering (ANNIE'95)
1415 North Dayton, St. Chicago, IL 60622 for providing the Conference, pp. 399 – 404, 1995.
complimentary copy of optimization software "Lingo11" for [13] Tolga Bektas, “The multiple traveling salesman
carrying out this study. problem: an overview of formulations and solution
procedures”, Omega The international journal of
management science, Vol. 34, pp. 209-219, 2006
7. REFERENCES www.elsevier.com/locate/omega
[1] Buthainah Fahran Al-Dulaimi, and Hamza A. Ali,”
[14] D. E. Goldberg, “Genetic Algorithm in Search,
Enhanced Traveling Salesman Problem Solving by
Optimization and Machine Learning”, Machine
Genetic Algorithm Technique (TSPGA)”, World
Learning. Addison-Wesley, New York, 1989.
Academy of Science, Engineering and Technology, Vol.
[15] P. Larranaga, C. M. H. Kuijpers, R. H. Murga, I. Inza
38, pp. 296-302, 2008.
and S. Dizdarevic, “Genetic Algorithms for the
[2] Moustapha Diaby, “The Traveling Salesman Problem:
Travelling Salesman Problem: A Review of
A Linear Programming Formulation”, Wseas
Transactions On Mathematics, Vol. 6, Issue 6. June Representations and Operators”, Artificial Intelligence
2007. Review. 13, pp. 129-170, 1999.
[16] S. R. Shubhra, B. Sanghamitra and K. Pal. Sankar,
[3] Zakir H. Ahmed, “Genetic Algorithm for the Traveling
“New Operators of Genetic Algorithms for Traveling
Salesman Problem using Sequential Constructive
Salesman Problem”, IEEE, 0-7695-2128-2/04, 2004.
Crossover Operator”, International Journal of
[17] Ji. Ping and Ho. William, “The Traveling Salesman and
Biometrics & Bioinformatics (IJBB) Vol.3, No.6, pp.
the Quadratic Assignment Proble: Integration, Modeling
97-105.
[4] I. Choi, S. Kim and H. Kim. "A genetic algorithm with and Genetic Algorithm”, Int. Symposium on OR and its
a mixed region search for the asymmetric traveling Applications, pp.198-205, 2005
[18] C. Ding, Ye. Cheng and M. He, “Two-Level Genetic
salesman problem". Computers & Operations Research
Algorithm for Clustered Traveling Salesman Problem
30, pp. 773 – 786, 2003.
with Application in Large Scale TSPs”, Tsinghua
[5] C. Moon, J. Kim, G. Choi and Y. Seo. "An efficient
Science and Technology, Vol.12, No.4, pp. 459- 465,
genetic algorithm for the traveling salesman problem
2007
with precedence constraints". European Journal of
[19] P. Borovska, T. Ivanova and H. Salem, “Efficient
Operational Research, Vol. 140, pp. 606- 617, 2002.
[6] D.E. Goldberg. "Genetic Algorithms in Search, Parallel Computation of the Traveling Salesman
Problem on Multicomputer Platform”, Proceedings of
Optimization, and Machine Learning". Addison-
the International Scientific Conference „Computer
Wesley, New York, 1989.
[7] C. Malandraki, M.S. Daskin, Time dependent vehicle Science‟ 2004, Sofia, Bulgaria, pp. 74-79, Dec. 2004.
routing problems: formulations, properties and heuristic
algorithms, J. Trans. Sci., Vol. 26, pp. 185–200, 1992.
[8] C. Malandraki, R.B. Dial, A restricted dynamic
programming heuristic algorithm for the time dependent
traveling salesman problem, European J. Oper. Res.,
Vol. 90, pp. 45–55, 1996.

169 | P a g e www.cirworld.com