Using an Hybrid Ant Algorithm to Solving the Frequency Assignment Problem in Cellular Mobile Telecommunication Systems

• Main Authors: Chih-Lung Chou, 周志隆
• Other Authors: Kung-Jeng Wang
• Format: Others
• Language: en_US
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/87888684572886710930

碩士 === 中原大學 === 工業工程研究所 === 91 === The demand for mobile telephone service is growing rapidly. On the other hand, the available electromagnetic frequency spectrum is severely limited. The aim of frequency assignment in the cellular mobile telecommunication system is to assign frequency channels to the radio transmitters in base stations in order to maximize some measures of spectral efficiency. The frequency assignment problem is known to be an NP-hard optimization problem. Thus, it is impractical to use the exact algorithms for realistic large sized problems. Most of the heuristic algorithms presented during the 1970’s to early 1990’s were categorized as the non-iterative approaches, which is based on the graph-coloring algorithms. Since 1990, different iterative approaches have been proposed for solving various versions of frequency assignment problem, such as the constraint satisfaction, artificial neural networks, simulated annealing, tabu search, genetic algorithm, etc. However, both kinds of solution approaches have its advantage and disadvantage. In this thesis we present the designing and implement of a hybrid heuristic algorithm, which is the ant-based algorithm conjunction with one of the sequential assignment strategies (i.e., the frequency-exhaustive assignment) to compute the minimum span frequency assignment problem. The performance is verified through solving two sets of well-known Philadelphia benchmark test problems in the literature, called the easy problem and the difficult problem, respectively. For the easy problems, our algorithm is easy to find optimum solutions for all eight variants of the given test problem. For difficult problems, though our method did not reach optimum solutions (lower bounds) in limited time, in some cases, we get better solutions than some existing algorithm. And, for each of these harder problems, we always quickly obtained a solution within average 6% deviation of the optimum, which can be considered a promising result.