Spectrum Allocation Algorithm Using Intelligent Optimization

Cognitive radio is an effective method to solve the problem of low spectrum efficiency. Spectrum allocation is a key technique in CR, which affects the rationality and effectiveness of spectrum sharing. The problem of graph-theory based spectrum allocation is known to be NP-hard, and intelligent algorithm is an effective method to solve them. Two kinds of efficient intelligent algorithms are proposed in this thesis to improve the spectrum efficiency, which reduce the time consumption and attain fairness simultaneously.Because of its good performance, Discrete Artificial Bee Colony(DABC) algorithm has been used in spectrum allocation. But its optimization performance is weakened by its disadvantages such as the premature, and lacking orientation. To solve this from spectrum allocation’s perspective, some improvements have been done as follows:(1)After analyzing the strategy of one-dimensional updating in DABC algorithm, the two-stage updating strategy is introduced in the Multi-strategy Discrete Artificial Bee Colony based Spectrum Allocation(MDABC-SA) algorithm. The multi-dimensional updating strategy is used in initial searches to find a better initial population rapidly, and the one-dimensional updating is used in later searches to perform fine line search;(2)Randomized selection is used in the updating dimension of DBAC, but the more ‘1’ in the solution, the higher network utilization can be achieved in spectrum allocation. So the strategy to update the elements with value only 0 is proposed to enhance the directionality and effectiveness.Differential Evolution has the advantages of fast convergence speed and highly flexible, but its optimization performance is weakened by some disadvantages in spectrum allocation, some improvements have been done as follows:(1)In order to solve the deficiency of exploitation capability in DE algorithm, the mutation operator which learn from the best solution in current iteration is proposed in this thesis;(2)Because the DE algorithm can not be used in discrete domain directly, the binary coding with the pro 1 feature is introduced in the Efficient and Adaptive Differential Evolution based Spectrum Allocation(EADE-MSA) algorithm, to enhance the directionality and effectiveness;(3)Well adapted to the search requirements of population, and balance both the exploration and the exploitation, an adaptive crossover operator, which is non linear changed with iterations is proposed in this thesis. The adaptive crossover operator make the DE algorithm has a better global search capability in initial searches, and a better local search capability in later searches, so the convergence speed and efficiency can be improved.The two proposed spectrum allocation algorithms are compared with other algorithms, the experimental results indicates that they all have an excellent performance in terms of convergence speed and solution accuracy. Among them, combining with the multi-objective utilization function, EADE-MSA can balance the spectrum utilization with fairness more efficiently in a shorter time.