Department of Management Engineering / Operations Research Department of Management Engineering / Operations Research

The Symmetric TSP

P
min d x
Branch and Bound for TSP P(i,j)E ij ij
s.t. x =2 i {1, 2, . . . , n}
Using the 1-tree relaxation Pj ij
i,jZ xij |Z| 1 Z V
Jesper Larsen and Jens Clausen xij 0, 1 (i, j) E
jla,jc@imm.dtu.dk

Department of Management Engineering

Technical University of Denmark

Jesper Larsen and Jens Clausen 1 Jesper Larsen and Jens Clausen 2

Department of Management Engineering / Operations Research Department of Management Engineering / Operations Research

Bounds The 1-tree bound

One way to identify a bound for the TSP is by
relaxing constraints. This could be to allow
subtours. This bound is although known to be
rather weak.
An alternative is the 1-tree relaxation.
Identify a special vertex 1 (this can be any
vertex of the graph).
1 and all edges incident with 1 are removed
from G.
For the remaining graph determine the
minimum spanning tree T .
Now the two smallest edges e1 and e2 incident
with 1 are added to T producing T1 (called a
1-tree)

Jesper Larsen and Jens Clausen 3 Jesper Larsen and Jens Clausen 4
Department of Management Engineering / Operations Research Department of Management Engineering / Operations Research

Why is T1 a bound?
We need to convince ourselves that the total cost of So the cost of T1 is less that or equal to the cost
T1 is a lower bound of the value of an optimal tour. of any Hamiltonian tour.
Note that a Hamiltonian tour can be divided into In the case T1 is a tour we have found the
two edges e1 and e2 that are incident with 1 and optimal solution and can prune by bounding.
the rest of the tour (let us call it T ). otherwise we need to bound.
So the set of Hamiltonian tours is a subset of
1-trees of G.
Since e1 , e2 are the two smallest edges incident
to 1 de1 + de2 de1 + de2 . Furthermore as T is a
tree d(T ) d(T ).

Jesper Larsen and Jens Clausen 5 Jesper Larsen and Jens Clausen 6

Department of Management Engineering / Operations Research Department of Management Engineering / Operations Research

TSP of Bornholm 1-tree bound of Bornholm

C C
A B C D E F G H
A
D 0 11 24 25 30 29 15 15 D
Tree in rest of G
B 11 0 13 20 32 37 17 17
B B

H
C 24 13 0 16 30 39 29 22 Edge left out by Kruskals MST algorithm
H
E D 25 20 16 0 15 23 18 12 E
A
A
1tree edge
E 30 32 30 15 0 9 23 15

G F F 29 37 39 23 9 0 14 21
G F
Cost of 1tree = 97
G 15 17 29 18 23 14 0 7
H
15 17 22 12 15 21 7 0

Jesper Larsen and Jens Clausen 7 Jesper Larsen and Jens Clausen 8
Department of Management Engineering / Operations Research Department of Management Engineering / Operations Research

Strengthening the bound Strengthen the bound

Idea: Vertices of T1 with high degree are C
Modified distance matrix:

A B C D E F G H
incident with too many attractive edges. A
0 11 24 25 29 29 16 15
Vertices of degree 1 have on the otherhand too D

B 11 0 13 20 31 37 18 17
many unattractive edges. B

30
H C 24 13 0 16 29 39 22
Define i as the degree of vertex i minus 2. E
D 25 20 16 0 14 23 19 12
P
A

E 29 31 29 14 0 8 23 14
Note that iV i equals 0 since T1 has n G F F 29 37 39 23 8 0 15 21
edges and therefore the degree sum is 2n. G 16 18 30 19 23 15 0 8

For each edge (i, j) E we transform the cost Cost of 1tree = 97

H
15 17 22 12 14 21 8 0

to dij = dij + i + j .

Jesper Larsen and Jens Clausen 9 Jesper Larsen and Jens Clausen 10

Department of Management Engineering / Operations Research Department of Management Engineering / Operations Research

How do we branch? Branching on Bornholm

Observe that in the case our 1-tree is not a tour
C

at least one vertex has degree 3 or more. B

H
E
A

So choose a vertex v with degree 3 or more. G F

For each edge (ui , v) generate a subproblem

Exclude (A,H) Exclude (H,G)
where (ui , v) is excluded from the set of edges. Exclude (D,H)

C C
C

D D
D

B B
B

H H
H
E E
E
A A
A

G F G F
G F

Cost of 1tree = 98 The optimal tour ! Cost of 1tree = 104

Jesper Larsen and Jens Clausen 11 Jesper Larsen and Jens Clausen 12