Designing Reliable Communication Networks with a Genetic Algorithm Using a Repair Heuristic

Abstract

This paper investigates GA approaches for solving the reliable communication network design problem. For solving this problem a network with minimum cost must be found that satisfies a given network reliability constraint. To consider the additional reliability constraint different approaches are possible. We show that existing approaches using penalty functions can result in invalid solutions and are therefore not appropriate for solving this problem. To overcome these problems we present a repair heuristic, which is based on the number of spanning trees in a network. This heuristic always generates a valid solution, which when compared to a greedy cheapest repair heuristic shows that the new approach finds better solutions with less computational effort.

Applications of the Cross-Entropy Method in Reliability

“…Yeh et al [33] proposed a method based on a Genetic Algorithm to optimize the k-node set reliability subject to a specified capacity constraint. Reichelt et al used a Genetic Algorithm in combination with a repair heuristic to minimize the network cost with specified network reliability constraints [25].…”

A Study of an Iterated Local Search on the Reliable Communication Networks Design Problem

“…In [6], Dengiz and Alabap introduce a simulated annealing algorithm. In [8], Reichelt et al propose a genetic algorithm using a repair heuristic. Baran et al [12] investigate topology design by a GA with multiple objectives.…”

“…The RRLS and ILS are implemented in C++. The GA uses the repair heuristic from [8] and is implemented in C++ using the GALib [17]. For the GA, we use a population size of 100, 50% replacement, a uniform crossover with a crossover probability of p cross =0.9 and a mutation probability of p mut =0.01.…”

“…as genetic algorithms (GA) proposed in [4,9,10,8] perform a high number of reliability evaluations in each generation. In this paper, we investigate a random restart local search (RRLS) and an iterated local search (ILS).…”

Reliable Communication Network Design With Evolutionary Algorithms

“…[12,24] In the past, heuristic optimization methods and especially evolutionary algorithms (EA) have already been applied to the RCND problem. [7,4,16,8,21] This paper presents two new EA-based approaches for the RCND problem and compares their performance to existing approaches. Both new approaches (LaBORNet and BaBORNet) use repair strategies that ensure that the allterminal reliability of the resulting network design is above some pre-defined threshold.…”