Optimization of telecommunication networks via genetic algorithms

Abstract

This thesis presents five different optimization problems using a proposed searching technique, Genetic Algorithms (GA). In order to show the success and effective searching ability of GA, we also use different searching techniques for comparing their results in each of these problems. In the first problem, we deal with minimization of multiple access interference of users for asynchronous Direct Sequence Code Division Multiple Access (DSICDMA) systems through selection of a subset of Gold codes from its family set. Two searching tools, Simulated Annealing (SA) and exhaustive-like search are adopted to compare with GA. In the second problem, minimization of CDMA hard handoff in the Personal Communication Services (PCS) network is tackled. It involves selection of some cells from a fixed number of cells for installing Base Station Controllers (BSCs) and PCS exchanges (PCXs). Another searching technique, SA, is utilized for comparison. In the third problem, minimization of the sum of total installation cost of BSCs and PCXs, weighted-distance connection cost between Base Stations and BSCs and between BSCs and PCXs and CDMA hard handoff cost in the PCS network is discussed. The selection method is similar to the one used in the second problem and Hierarchical GA is adopted for comparing with SA. Besides, in the fourth problem, we tackle minimization of the sum of total radio repeater cost and link cost in Wireless Local Loop (WLL) systems by choosing installation locations of radio repeaters from a fixed number of grids. The searching technique, Lagrangian Heuristics, is used for comparison. Finally, in the fifth problem, we deal with minimization of total capacity usage (total working capacity usage plus total backup capacity usage) in a self-healing Asynchronous Transfer Mode (ATM) network. It involves selection of a working virtual path and backup virtual path for each pair of source and destination nodes of multicast circuits. GA and exhaustive search are adopted to find results in seven networks for comparison of the three proposed methods, link restoration scheme using exhaustive search, path restoration scheme using exhaustive search and path restoration scheme using GA. On the whole, from the obtained results of the above five problems, GA outperforms other searching techniques in the single objective approach and enables to find Pareto tradeoff solutions along the tradeoff surface in the multiobjective approach. Therefore, the application of GA is successful in various optimization problems.

Date of Award	3 Oct 2001
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Kim Fung MAN (Supervisor)