Publication detail
Solving the Integer Maximal Multicommodity Flow Problem Using Simulated Annealing
ŠEDA, M.
English title
Solving the Integer Maximal Multicommodity Flow Problem Using Simulated Annealing
Type
Chapter in a book
Language
en
Original abstract
In this paper the Integer Maximal Multicommodity Flow Problem is discussed. Multicommodity flow problems have many specific formulations depending on the constraints defined resulting in various applications in transportation, distribution and telecommunications. Since its integer version belongs to the class of NP-hard combinatorial problems, for large scale instances, it must be solved by approximation or heuristic techniques. We present a stochastic heuristic approach based on a simulated annealing algorithm. All evaluations of the objective function in this algorithm are provided by an allocation procedure. Since the allocation of the edge capacities among the commodities and the corresponding combined maximal flow depend on the order in which the commodities are selected, the neighbouring mechanism in simulated annealing is set to generate permutations of commodities. Computational results show that, for suitable parameter settings presented in the paper, this approach is able to find the optimal solution almost in all cases or at least a solution very close to optimum when the test is executed several times. Furthermore, the proposed algorithm is stable because even the average results gained from all the executions are close to optimum.
Released
2004-10-01
Publisher
DAAAM International Wien
Location
Wien
ISBN
3-901509-38-0
ISSN
1726-9687
Book
Katalinic, B. (ed.): DAAAM International Scientific Book 2004
Pages from–to
553–
Pages count
10
BIBTEX
@inbook{BUT55521,
author="Miloš {Šeda}",
title="Solving the Integer Maximal Multicommodity Flow Problem Using Simulated Annealing",
booktitle="Katalinic, B. (ed.): DAAAM International Scientific Book 2004",
year="2004",
publisher="DAAAM International Wien",
address="Wien",
series="DAAAM International Scientific Book 2004",
pages="10",
isbn="3-901509-38-0"
}