A study on resource management in communication networks by heuristic approach

Abstract

Both present and future communication networks are expected to support multiple and diverse applications and meet various quality of service (QoS) requirements. Accordingly, a key issue in the design and operation of a network is how the related resources should be provided in order to meet the requirements of each connection. Thus, this thesis addresses three kinds of network optimization problems for resource management in communication networks from the viewpoint of efficiently and effectively constructing and managing the network using limited resources In order to solve the problems considered in this thesis, we introduce the following evolutionary computation algorithms-the genetic algorithm (GA) and coevolutionary algorithm (Co-EA)- as well as the tabu search (TS), all of which have proven useful in solving combinatorial problems. In applying the proposed methods to each problem, many of the elements are contrived to improve solution quality and computational efficiency, such as the genetic representation, evaluation function, genetic operators and procedures for GA and Co-EA; and the neighborhood structure and search strategy for TS. To promote population diversity and search efficiency in the algorithm, we also adopt strategies of localized evolution and steady-state reproduction, and develop the methods of selecting environmental individuals and evaluating fitness for Co-EA. In the first problem, we propose heuristic evolutionary computation algorithms that can simultaneously solve the route selection and rate allocation problem in multirate multicast networks; that is, the problem of constructing multiple multicast trees and simultaneously allocating the rate of receivers f01 maximizing the sum of utilities over all receivers, subject to link capacity and delay constraints tor high-band width delay-sensitive applications in point-to point Communication networks. The results of extensive computational simulations show that the proposed algorithms...