Exponential Weighting based Network Ranking for Heterogeneous Wireless Networks

Abstract

Heterogeneous wireless networks (HWNs) provide communication related services in some specific region using multiple wireless access networks so that the users can access the networks with better quality of service (QoS). Typical component level access networks include wireless local area networks (WLAN), Worldwide Interoperability for Microwave Access (WiMAX), 2G, 3G and 4G networks. This means that when a user terminal moves in HWNs environment, it experiences different available networks that must be ranked according to some criteria before deciding which network is suitable for the user terminal. This also relates to the activity/business of the user terminals for which they require the access of network, that is streaming, conversation, interactive and background use. The problem of ranking and selecting the best suitable network among multiple access networks according to some criteria that fulfils the user requirements and network performance attributes is an active research area in HWNs. Multiple network selection algorithms, including analytical hierarchy process (AHP), technique for order preference by similarity to ideal solution (TOPSIS), utility theory, multiplicative exponent weighting (MEW), and simple additive weighting (SAW) have been developed to handle this challenge. Their applications in some cases include the network performance attributes, while others involve user preference without using inherited network attributes. Therefore, the network selection algorithms are often integrated to cover both the user preferences and the network performance attributes. In this thesis, we utilized the AHP and TOPSIS algorithm and integrated their associated weights through multiplicative exponential weighting (MEW) approach. This allows us to cover both the user preferences and the network attributes in the network selection process. The AHP method identify a weighting criterion for the user preference and TOPSIS select weights based on current network attributes. The integration of the two classes of weighting criteria can be given equal or any other weightage in network ranking. The proposed approach of utilizing multiplicative exponent weighting in the use of network selection is compared with the existing multiplicative weighting by considering three predefined scenarios. It is observed that the proposed method shows better, or equivalent results as compared to the existing approach in all the three scenarios.