A Self-adaptive Access Selection Algorithm for Cognitive Networks Based on Fuzzy Neural Networks

Abstract

In order to improve the performance of access network and terminal services, an intelligent access selection algorithm based on fuzzy neural networks is presented. By establishing fuzzy neural networks that meets specific needs, network bandwidth, delay, load balancing at network side and moving speed at terminal-side are considered as access factors determined the pros and cons for access network. Then, fuzzification of access factors above by fuzzy logic is done, and fuzzy neural networks are employed to inference. At last, the optimal access network can be selected. To improve the performance of parameters learning, a fuzzy particle swarm optimization algorithm is proposed. Compared to the basic particle swarm optimization algorithm, fuzzy particle swarm optimization can overcome the disadvantages that convergence speed slows down even tends to stagnate in the latter, ensure parameter optimization quality for fuzzy neural networks. Experimental results show the intelligent access selection algorithm can reduce terminal access blocking rate and packet loss rate, as well as increase the average throughput of access network effectively.