Network Selection In Heterogeneous Wireless Networks Based On Multiple Attributes Decision Making

The competition and development of mobile communications and broadband wireless access technologies make future communication networks heterogeneity, and each network will experience the evolution of isolation, interoperability and integration. In the framework of heterogeneous wireless networks, an important job is to have research on network selection algorithm. Selecting the optimal access network makes full use of the complementary characteristics of different wireless access technologyies in coverage, mobility support and QoS performance etc., and then achieves multi-Radio Access Gain, guarantees QoS and optimizes system resources regulation.In the next generation network, users’ service need become richer. Based on different QoS requirements for each class of service defined by3GPP, the decision attributes weight for each class of service is calculated by Analytic Hierarchy Process(AHP). SAW(Simple Additive Weight), MEW(Multiplicative Exponent Weighting), TOPSIS(Technique for Order Preference by Similarity to an Ideal Solution), GRA(Grey Relational Analysis) and Grey Projection Analysis algorithms are used with weight for determining the optimal access network. These five network selection algorithms are simulated and analyzed comparatively, and the results show that TOPSIS and GRA algorithms which are superior to MEW, SAW and Grey Projection Analysis algorithms in weight sensitivity, can balance the allocation of four classes of service, to avoid a single network overload.Heterogeneous network selection decision depends on user preferences, network performance and terminal performance. According to fuzzy decision theory, a fuzzy multiple attributes decision algorithm is used for network selection. First, subjective and objective decision attributes weights of each available network are calculated by fuzzy AHP and entropy method respectively. Game Theory is used to get combination weights, which takes both users’ subjective preferences and network objective information into account. Meanwhile, the utility functions of decision attributes are built based on diverse QoS elasticities and real-time requirements of different services. At last, each available network’s evaluation value is obtained by fuzzy RTOPSIS with combination weights. Simulation shows that the algorithm realizes personalized network selection.As to load imbalance and dynamic changes in heterogeneous network, multi-attribute fuzzy neural network method is applied to network selection, presenting a load balance algorithm. On one hand, by applying multi-attribute pre-decision, the algorithm decreases fuzzy rules, reduces computational complexity and guarantees users’ QoS; On the other hand, the algorithm executes reinforcement learning with the objective of the equal resource utilization of available networks, adjusts the membership function parameters of fuzzy neural network by error back propagation and then achieves access judgment intelligently. The simulation results show that the algorithm balances the load of heterogeneous systems, decreases blocking probability as well as guarantees QoS.