The utility based non-linear fuzzy AHP optimization model for network selection in heterogeneous wireless networks

Highlights

• The disadvantages of using extent analysis method for network selection problems are discussed.

• Weights of network parameters are obtained by applying a nonlinear fuzzy optimization model.

• Consistency Index with this proposed model is better than the existing non-linear models.

• Parameterized utility functions are used to evaluate the utility values of network attributes.

• Results obtained for network selection with the MEW method are better than TOPSIS and SAW methods.

Abstract

Next generation wireless networks will integrate various heterogeneous technologies like WLAN, WiMax and cellular technologies etc., to support multimedia services with higher bandwidth and guaranteed quality of service (QoS). In order to keep the mobile user always connected to the best wireless network in terms of QoS parameters and user preferences, an optimal network selection technique in heterogeneous networks is required. This paper proposes a novel fuzzy-Analytic Hierarchy Process (AHP) based network selection in heterogeneous wireless networks. Triangular fuzzy numbers are used to represent the elements in the comparison matrices for voice, video and best effort applications. Deriving crisp weights from these fuzzy comparison matrices is a challenging task. When extent analysis method is applied, irrational zero weights are obtained for some attributes. Due to this, many important criteria are not considered in the decision making process. To overcome this problem, a new non-linear fuzzy optimization model for deriving crisp weights from fuzzy comparison matrices for network selection is presented. The weights obtained from this model are more consistent than the existing optimization models. Also, parameterized utility functions are used to model the different Quality of Service (QoS) attributes (bandwidth, delay, jitter, bit error rate) and user preferences (cost) for three different types of applications. Finally, scores are calculated exclusively for each network by three MADM (Multiple Attribute Decision Making) methods Simple Additive Weighting (SAW), TOPSIS (The Technique for Order of Preference by Similarity to Ideal Solution) and MEW (Multiplicative Exponential Weighting). Results show that the MEW method gives more appropriate scores with utility functions than the SAW and TOPSIS methods.

Introduction

The various heterogeneous wireless technologies like WLAN, WiMax and cellular technologies like 2G/3G etc., will be integrated in future generation wireless networks through a single IP-based core. When a mobile device roams in these heterogeneous environments, it undergoes vertical and horizontal handovers continuously. In order to provide Always Best Connected (ABC) property, an optimal Vertical Handover Decision (VHD) is required. Many VHD methods are presented in the literature that includes fuzzy logic, game theory, function based, user-centric and Markov decision process [1]. MADM methods include multiple attributes with medium complexities while the other methods have higher complexity with the increase in number of attributes. So, in this paper, the focus is only on MADM based network selection methods.

The earlier study [2] has presented various MADM methods. Among these, most popular methods are SAW [3] MEW [4], TOPSIS [5] PROMETHEE [6] and ELECTRE [7]. AHP method [8] is widely used for providing weights to these MADM methods. Song and Jamalipour [9] have presented AHP and Grey Rational Analysis (GRA) based technique where AHP has been used for selection criteria and GRA evaluates heterogeneous networks. The proposed methodology combines the AHP to decide the relative weights of criteria set according to network’s performance, as well as the GRA to rank the network alternatives. Goyal and Kaushal [10] have applied AHP method to find the effect of mobility of the mobile device on the handover decision process in heterogeneous wireless networks. Lahby et al. [11] have proposed an enhanced TOPSIS method by using the Analytic Network Process (ANP) to weight the criteria and then apply the TOPSIS method. Various methods have been developed for the weight calculation of network attributes. Ahuja et al. [12], [13] have proposed a novel network selection algorithm in a typical heterogeneous environment of EDGE and UMTS networks where the selection metrics are combined with PSO for relative dynamic weight optimization. Cost functions are used for network selection. Charilas et al. [14] have presented a framework that defines the weights of attributes based on the variance of network measurements. Principal Component Analysis and AHP are applied to derive parameter weights through pairwise comparison matrices. Yang and Tseng [15] have proposed handoff decision algorithm that combines Weighted Rating Mean Analysis (WRMA) to rate the relative importance of network attributes and a fuzzy methodology of TOPSIS is used to select candidate networks. Ahuja et al. [16] have proposed a network selection algorithm in which weight estimation of attributes is computed using entropy and TOPSIS method has been used for network ranking. To achieve the consistency ratio of greater than 0.1 in AHP, Chandavarkar and Guddeti [17] have proposed a Simplified and Improved AHP (SI-AHP) technique which automatically computes the reciprocal matrix of attributes. This technique is further used in Simplified and Improved Multiple Attributes Alternate Ranking method (SI-MAAR) by Chandavarkar and Guddeti [18] for solving the rank reversal problem when a network is added or removed from the selection list. In many situations, decision makers have to model their preferences with fuzzy values. But, all these approaches have not considered the fuzzy weights for modelling the AHP matrix.

Fuzzy-based MADM methods are integrated for different types of selection and ranking problems [19], [20], [21], [22], [23]. Different fuzzy based MADM methods have also been used for vertical handovers [24]. Chamodrakas and Martakos [25] have presented a novel method in which diverse QoS elasticises are modeled by parameterized utility functions and the fuzzy set representation of TOPSIS method is used to rank the networks. Mehbodniya et al. [26] have presented Fuzzy Logic Controllers (FLCs) with on Fuzzy VIKOR (FVIKOR) to select the networks in heterogeneous wireless networks. Abid et al. [27] propose an innovative handover decision making scheme based on a single criterion utility function that rates user satisfaction and captures sensitivity for each decision criterion. Chandavarkar and Guddeti [28] have compared the performances of different sigmodial utility functions in heterogeneous wireless networks. Drissi et al. [29] proposes a context aware network selection based on utility functions that consider QoS requirements and user preferences. Kosmides et al. [30] introduce three utility functions based on the type of application that users request. Drissi et al. [31] have used Fuzzy-AHP with SAW method to keep the users Always Best Connected (ABC). Khan et al. [32] have used fuzzy rule based system to eliminate inappropriate networks and the selection of networks is done by TOPSIS method. Zineb et al. [33] have proposed an enhanced vertical handover based on Fuzzy Inference MADM approach for heterogeneous networks. This method is based on a MADM technique with fuzzy logic inference system to reduce handover decision time. Skondras et al. [34] have applied ANP to estimate the weights, and Fuzzy TOPSIS method has been implemented for network selection. But these methods have not applied any fuzzy prioritization methods to derive crisp weights from fuzzy comparison matrices.

Chang [35] has proposed an extent analysis method to Fuzzy-AHP comparison matrices for deriving crisp weights. Charilas et al. [36] have applied the extent analysis method on Fuzzy-AHP to derive crisp weights of the network attributes, and ELECTRE method is used to rank the wireless networks. Also, Brajković et al. [37] have proposed an extent analysis Fuzzy-AHP with TOPSIS method for optimal wireless networks. Goyal et al. [38] have also used extent analysis method for network selection in heterogeneous wireless networks considering various network parameters. Goyal and Kaushal [39] have analyzed the effect of utility functions on the extent analysis method for vertical handover scenarios. But, extent analysis method can lead to irrational zero weights in many cases ([40]). Due to this, many important criterias are not considered in the decision making process. Therefore, extent analysis method can lead to zero weights for important network parameters in case of vertical handover decision problems also.

In this paper, a novel fuzzy-Analytic Hierarchy Process (AHP) based network selection technique is presented in heterogeneous wireless networks for voice, video, and best-effort applications. Triangular fuzzy numbers are used to represent the elements in the comparison matrices. When extent analysis method is applied, irrational zero weights are obtained for some attributes in all the three network applications. To overcome this problem, a new non- linear fuzzy optimization model for deriving crisp weights from fuzzy comparison matrices is proposed. Consistency of the comparison matrix is important as it determines how consistent the judgements are made by the decision maker in the comparison matrix. Inconsistencies may typically arise when many pairwise comparisons are performed. The weights obtained from the proposed model are consistent than other fuzzy optimization methods. Also, parameterized utility functions are used to model the different Quality of Service (QoS) attributes for different applications with SAW, TOPSIS and MEW methods. With the utility functions, the actual utility values of network parameters are considered for the network selection. Therefore, handover decisions are more appropriate with the proposed utility based network selection method.

Rest of the paper is organized as follows. Section 2 presents the preliminaries required for the MADM based network selection methods. In section 3, the proposed Fuzzy-AHP based MADM technique for network selection is discussed. Section 4 presents the results and analysis of the proposed model with SAW, TOPSIS and MEW methods. Finally, Section 5 concludes with future directions of research.