Solving the delay-constrained capacitated minimum spanning tree problem using a dandelion-encoded evolutionary algorithm

Abstract

The Delay-Constrained Capacitated Minimum Spanning Tree (DC-CMST) is a recently proposed problem which arises in the design of the topology of communications networks. The DC-CMST proposes the joint optimization of the network topology in terms of the traffic capacity and its mean time delay. In this paper, an evolutionary algorithm which uses Dandelion-encoding is proposed to solve the problem. The Dandelion code has been recently proposed as an effective way of encoding trees in evolutionary algorithms, due to its good properties of locality. We describe the main characteristics of the algorithm, and compare its results with that of an existing heuristic for the DC-CMST. We show that our Dandelion-encoded evolutionary algorithm is able to obtain better results in all the instances tackled. © 2008 Springer Berlin Heidelberg.