Evolutionary algorithms for orthogonal frequency division multiplexing-based dynamic spectrum access systems

摘要

This paper proposes two evolutionary algorithms (EAs) to perform dynamic spectrum assignment in distributed OFDM-based cognitive radio access networks. To achieve better radio resource utilization, the central spectrum manager (CSM) jointly considers the type of modulation level which can be used by each radio pair when deciding the number of subcarriers to be assigned. Using the piecewise convex transformations, we reformulate the nonlinear integer programming problem to an integer linear programming so that it is possible to obtain the optimal solution. While the solution to the integer linear programming problem is NP-hard, the proposed EAs provide useful suboptimal solutions especially when the number of radios and subcarriers are large. Our first proposed EA adopts the genetic algorithm. Although the reproduction process generates chromosomes which do not fulfill the constraints, our algorithm integrates the invisible walls technique used in particle swam optimization to retain the diversity while constructing chromosomes for the next generation. The second EA adopts the ant colony optimization approach using a directed multigraph. The vertices are used to represent the subcarriers and each edge corresponds to a possible chosen modulation index of a specific radio. We further obtain the performance of the two EAs through simulations and benchmark them against the optimal solution. Our studies show that ant colony algorithm gives better solutions most of the time, however, its computation time increases much faster compared to generic algorithm when the numbers of users and subcarriers increase.