Travelling Salesman Problem using Branch and

Bound Approach
Chaitanya Pothineni
December 13, 2013

Abstract
To find the shortest path for a tour using Branch and Bound for finding
the optimal solutions.A branch-and-bound algorithm consists of a system-
atic enumeration of all candidate solutions, where large subsets of fruitless
candidates are discarded ,by using upper and lower estimated bounds of
the quantity being optimized.

1 Introduction
The Traveling Salesman Problem, TSP for short, deals with creating the ideal
path that a salesman would take while traveling between cities. The solution
to any given TSP would be the cheapest way to visit a finite number of cities,

1
visiting each city only once, and then returning to the starting point. We also
must assume that if there are two cities, city A and city B for example, it costs
the same amount of money to travel from A to B as it does from B to A.
For the most part, the solving of a TSP is no longer executed for the intention
its name indicates. Instead, it is a foundation for studying general methods that
are applied to a wide range of optimization problems.

1.1 Applictions
The TSP naturally arises as a subproblem in many transportation and
logistics applications.
Scheduling of a machine to drill holes in a circuit board or other object.
In this case the holes to be drilled are the cities, and the cost of travel is the
time it takes to move the drill head from one hole to the next.

2 History
The problem was first formulated in 1930 and is one of the most intensively
studied problems in optimization. It is used as a benchmark for many optimiza-
tion methods. Even though the problem is computationally difficult, a large
number of heuristics and exact methods are known, so that some instances with
tens of thousands of cities can be solved.
The Traveling Salesman Problem (TSP) is a problem whose solution has
eluded many mathematicians for years. Currently there is no solution to the
TSP that has satisfied mathematicians. Historically, mathematics related to the
TSP was developed in the 1800’s by Sir William Rowan Hamilton and Thomas
Penyngton Kirkman, Irish and British mathematicians,respectively. Hamilton
was the creator of the Icosian Game in 1857. It was a pegboard with twenty
holes that required each vertex to be visited only once, no edge to be visited
more than once, and the ending point being the same as the starting point.This
kind of path was eventually referred to as a Hamiltonian circuit.However, the
general form of the TSP was first studied by Karl Menger in Vienna and Harvard
in the late 1920’s or early 1930’s.
TSP?s were first studied in the 1930?s by mathematician and economist Karl
Menger in Vienna and Harvard. It was later investigated by Hassler Whitney
and Merrill Flood at Princeton.
In 1994, Applegate, Bixby , Chvatal, and Cook solved TSP containing 7,397
cities. Later in 1998, they solved it using 13,509 cities in United States. In 2001,
Applegate, Bixby, Chvtal, and Cook found the optimal tour of 15,112 cities in
Germany. Later in 2004, TSP of visiting all 24,978 cities in Sweden was solved;
a tour of length of approximately 72,500 kilometers was found and it was proven
that no shorter tour exists. This is currently the largest solved TSP. [1]

2
3 Optimal Solution for TSP using Branch and
Bound
A branch-and-bound algorithm consists of a systematic enumeration of all can-
didate solutions, where large subsets of fruitless candidate s are discarded ,by
using upper and lower estimated bounds of the quantity being optimized.
The Branch and Bound strategy divides a problem to be solved into a number
of sub-problems. It is a system for solving a sequence of subproblems each of
which may have multiple possible solutions and where the solution chosen for
one sub-problem may affect the possible solutions of later sub-problems.
Suppose it is required to minimize an objective function. Suppose that we
have a method for getting a lower bound on the cost of any solution among
those in the set of solutions represented by some subset. If the best solution
found so far costs less than the lower bound for this subset, we need not explore
this subset at all.

Let S be some subset of solutions.

L(S)=a lower bound on the cost of any solution belonging to S
Let C=cost of the best solution found so far
If C ≤ L(S),there is no need to explore S because it does not contain any better
solution.
If C > L(S),then we need to explore S because it may contain a better solution.
[2]

4 Branch and Bound approach Algorithm

function CheckBounds(st,des,cost[n][n]) [3] . Cal. the bounds
Global variable: cost[N ][N ] - the cost assignment.
pencost[0] = t
for i ← 0, n − 1 do
for j ← 0, n − 1 do
reduced[i][j] = cost[i][j]
end for
end for
for j ← 0, n − 1 do
reduced[st][j] = ∞
end for
for i ← 0, n − 1 do
reduced[i][des] = ∞
end for
reduced[des][st] = ∞
RowReduction(reduced)
ColumnReduction(reduced)
pencost[des] = pencost[st] + row + col + cost[st][des]
return pencost[des]

3
end function
function RowMin(cost[n][n],i) . Cal. min in the row
min = cost[i][0]
for j ← 0, n − 1 do
if cost[i][j] < min then
min = cost[i][j]
end if
end for
return min
end function
function ColMin(cost[n][n],i) . Cal. min in the col
min = cost[0][j]
for i ← 0, n − 1 do
if cost[i][j] < min then
min = cost[i][j]
end if
end for
return min
end function
function Rowreduction(cost[n][n]) . makes row reduction
row = 0
for i ← 0, n − 1 do
rmin = rowmin(cost, i)
if rmin 6= ∞ then
row = row + rmin
end if
for j ← 0, n − 1 do
if cost[i][j] 6= ∞ then
cost[i][j] = cost[i][j] − rmin
end if
end for
end for
end function
function Columnreduction(cost[n][n]) . makes column reduction
col = 0
for j ← 0, n − 1 do
cmin = columnmin(cost, j)
if cmin 6= ∞ then
col = col + cmin
end if
for i ← 0, n − 1 do
if cost[i][j] 6= ∞ then
cost[i][j] = cost[i][j] − cmin
end if
end for

4
 end for
end function
function Main . main function
for i ← 0, n − 1 do
select[i] = 0
end for
rowreduction(cost)
columnreduction(cost)
t = row + col
while allvisited(select) 6= 1 do
for i ← 1, n − 1 do
if select[i] = 0 then
edgecost[i] = checkbounds(k, i, cost)
end if
end for
min = ∞
for i ← 1, n − 1 do
if select[i] = 0 then
if edgecost[i] < min then
min = edgecost[i]
k=i
end if
end if
end for
select[k] = 1
for p ← 1, n − 1 do
cost[j][p] = ∞
end for
for p ← 1, n − 1 do
cost[p][k] = ∞
end for
cost[k][j] = ∞
rowreduction(cost)
columnreduction(cost)
end while
end function

5
5 Evaluation of Algorithm
The following matrix is the Cost Matrix which shows the distance be-
tween the two cities.
1 2 3 4 5
 
1 − 10 8 9 7
2 
 10 − 10 5 6 

Cost Matrix = 3 
 8 10 − 8 9 

4  9 5 8 − 6 
5 7 6 9 6 −
We know that the sum of row minimum gives us the lower bound.Now
we have to find the reduced matrix by subtracting the minimum element from
every row.So,row minimum will be 31.
1 2 3 4 5
 
1 − 3 1 2 0
2 
 5 − 5 0 1 

Reduced Matrix = 3 
 0 2 − 0 1 

4  4 0 3 − 1 
5 1 0 3 0 −
In the above reduced matrix there should be a Zero in every row
and every column.Now apply the column minimum principle which means for a
particular city I will have come into that city from other city.So,now subtracting
the column minimum , we get the lower bound 31+1 = 32.
1 2 3 4 5
0(2) 0(1)
 
1 − 3 2
2 
 5 − 4 0(1) 1 

Reduced Matrix = 3 
 0(1) 2 − 0(0) 1 

4  4 01) 2 − 1 
5 1 0(0) 2 0(0) −
Now we have to make the assignments Zero’s in the above matrix.If
we can not make an assignment at that particular Zero we have to go for the
next highest element in that row.If there is same value,we have to calculate the
sum of row penalty and column penalty.If the sum of penalties is more then we
have to make that assignment because we incur an additional cost if we don’t
make that assignment.
Now if we don’t have path from 1 to 3 i.e,

X(13) = 0 (1)

we have an additional cost of 2 and the lower bound becomes 32+2 = 34.
If we have path from 1 to 3 i.e,

X(13) = 1 (2)

6
.
we eliminate the respective row and column.

1 2 4 5
 
2 5 − 0 1
3  − 2 0 1 
Reduced Matrix =  
 4 0 − 1 
4
5 1 0 0 −

We have to check whether every row and every column has a zero otherwise we
have to subtract the minimum element from every respctive row or column.So,we
get
1 2 4 5
 
2 4 − 0 0
3  − 2 0 0 
Reduced Matrix =  
 3 0 − 0 
4
5 0 0 0 −
Here we made a column reduction for two columns(1 & 5)which gives a lower
bound of 32+1+1 = 34.
Again we calculate the penalty for the above matrix.

1 2 4 5
0(0) 0(0)
 
2 4 −
3  − 2 0(0) 0(0) 
Reduced Matrix =  
4  3 0(0) − 0(0) 
5 0(3) 0(0) 0(0) −

Now if we don’t have path from 5 to 1 i.e,

X(51) = 0 (3)

we have an additional cost of 3 and the lower bound becomes 34+3 = 37.
If we have path from 5 to 1 i.e,

X(51) = 1 (4)

.
we eliminate the respective row and column.

2 4 5
 
2 − 0 0
Reduced Matrix = 3  2 0 − 
4 0 − 0

Here we a Zero in every row and column.So,the bound remains the same i.e,
34+0 = 34.

7
 If we calculate the penalty for the above matrix,

2 4 5
0(0) 0(0)
 
2 −
Reduced Matrix = 3  2 0(2) − 
4 0(2) − 0(0)

Now if we don’t have path from 3 to 4 i.e,

X(34) = 0 (5)

we have an additional cost of 2 and the lower bound becomes 34+2 = 36.
If we have path from 3 to 4 i.e,

X(34) = 1 (6)

.
we eliminate the respective row and column.

2 5
 
2 − 0
Reduced Matrix =
4 0 −

Here we a Zero in every row and column.So,the bound remains the same i.e,
34+0 = 34.
Now we have only two possible assignments with Zero penalty cost which
would give us a feasible solution.

X(25) = 1 (7)

X(42) = 1 (8)
On adding all the paths we get,

X(13) + X(34) + X(42) + X(25) + x(51) = 34 (9)

References
[1] http://en.wikipedia.org/wiki/Travelling salesman problem.
[2] http://lcm.csa.iisc.ernet.in/dsa/node187.html.

[3] http://www.youtube.com/watch?v=nN4K8xA8ShM, 2010.