You are on page 1of 15
SEMINAR PAPER at the University of Applied Sciences Technikum Wien Game Engineering and Simulation Scalability
SEMINAR PAPER
at the University of Applied Sciences Technikum Wien
Game Engineering and Simulation
Scalability of a parallel implementa-
tion of ant colony optimization
by
Emanuel Plochberger,BSc
3481, Fels am Wagram
Supervisor: Dipl.-Ing. Dr. Markus Schordan
Vienna, December 14, 2010

Abstract

Contents

1 Introduction

 

1

2 Ant Colony Optimization

2

 

 

3 Implementation

 

6

 

 

4 Implementation

 

9

 

 

5

Conclusion

 

11

Bibliography

12

1 Introduction

2 Ant Colony Optimization

2.1 Swarm intelligence

2.2 Ant Colony Optimization

2.2.1 Agents The main differences to real ants [1]: • • • •

2.2.1 Agents

The main differences to real ants [1]:

Behaviour of ants [5]

2.2.2 Algorithm

t = 0 τ ij = c m n

τ ij (t + n) τ ij

t = t + n τ ij

2.2.3 Formulas

Trail[1]

ij i j

τ ij (t) ij τ ij ρ

τ ij (t + n) = ρ τ ij (t) + ∆τ ij

τ ij

τ ij = Σ m

k=1 τ

k

ij

k

τ ij Q τ ij = 0

k

τ ij = Q

k

L

k

0

2.2.4 Path selection

η ij τ ij α β

k

p ij (t) =

[τ ij (t)] α [η ij ]

β

Σ k [τ ik (t)] α [η ik ] β

3 Implementation

3.1 Synchronous Parallel Processing Ants

3 Implementation 3.1 Synchronous Parallel Processing Ants

N

S asymptotic (m) =

T seq (m) T par (m,

) = O(m

)

3

O(m 2 ) = O(m)

T seq (m) = O(m 3 ) m

T par (m, ) = O(m 2 ) m

S(m, N ) =

O(m 3 )

O(m 3 /N ) + T ovh (m, N )

T ovh (m, N )

3.2 Partially Asynchronous Parallel Processing Ants

4 Implementation

4.1 Computational Experience

4.1.1 Parameters for ACO

ρ

α

β

4.1.2 Results

speedup = time to solve the problem with a single processor time to solve the problem with P processors

eff iciency = speedup

P

 

5 Conclusion

Bibliography