ABC-PSO for vertical handover in heterogeneous

ARTICLE IN PRESS

JID: NEUCOM

[m5G;March 16, 2017;4:31]

Neurocomputing 0 0 0 (2017) 1–19

Contents lists available at ScienceDirect

Neurocomputing journal homepage: www.elsevier.com/locate/neucom

ABC-PSO for vertical handover in heterogeneous wireless networks Shidrokh Goudarzi a, Wan Haslina Hassan a, Mohammad Hossein Anisi b,∗, Ahmad Soleymani c, Mehdi Sookhak d, Muhammad Khurram Khan e, Aisha-Hassan Abdalla Hashim f, Mahdi Zareei a a

Communication System and Network (iKohza) Research Group, Malaysia Japan International Institute of Technology (MJIIT), University Teknologi Malaysia, 81310 Skudai, Johor, Malaysia b School of Computing Sciences, University of East Anglia, Norwich Research Park, Norwich, Norfolk, NR4 7TJ, United Kingdom c Department of Computing, Faculty of Computing, University of Teknologi Malaysia, Johor Bahru, Malaysia d Department of Systems and Computer Engineering, Carleton University, Ottawa, Canada e Center of Excellence in Information Assurance (CoEIA), King Saud University, Riyadh 12372, Saudi Arabia f Faculty of Electrical and Computer Engineering, International Islamic University Malaysia Kuala Lumpur, Kuala Lumpur, Malaysia

a r t i c l e

i n f o

Article history: Received 7 June 2016 Revised 26 July 2016 Accepted 10 August 2016 Available online xxx Keywords: Vertical handover ABC PSO IEEE 802.21 Heterogeneous wireless networks

a b s t r a c t Cloud computing is currently emerging quickly as a client-server technology structure and, currently, providing distributed service applications. However, given the availability of a diverse range of wireless access technologies, people expect continuous connection to the most suitable technology that matches price affordability and performance goals. Among the main challenges of modern communication is the accessibility to wireless networks using mobile devices, with a high service quality (QoS) based on preferences of the users. Past literatures contain several heuristic approaches that use simplified rules to look for the best network that is available. Nevertheless, attributes of mobile devices need algorithms that are quick and effective in order to select best available network near real-time. This study proposes a hybrid intelligent handover decision algorithm primarily founded on two main heuristic algorithms: Artificial Bee Colony or ABC as well as Particle Swarm Optimization or PSO named ABC-PSO to select best wireless network during vertical handover process. The ABC-PSO algorithm has been optimized to achieve small cost function that are powered using the IEEE 802.21 standard taking into account different available wireless networks, the application requirements and the user preferences to improve QoS. Numerical results demonstrate that the ABC-PSO algorithm compared to the related work has lower cost and delay, higher available bandwidth and less number of handover. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Nowadays, the mobile cloud computing has been advanced with much attention worldwide. Generally, the mobile cloud computing is one of the most rapidly emerging future technologies [1–4]. Although mobile devices has limited capacity to store and execute processes, information properties query process is still considered as one of the main issues for implementing mobile cloud computing environment. Heterogeneous wireless networks architecture then utilized to address this major issue. Heterogeneous wireless networks merging to become a complete IP network in the move toward setting up the next-generation network. Various access technologies, in this paradigm, require interconnection, which means vertical handovers are a necessity for mobility

∗

Corresponding author. E-mail addresses: [email protected], [email protected] (M.H. Anisi).

that is seamless. Wireless applications, devices, and networks have developed tremendously in the past decades. Nevertheless, the diverse methods and objectives deriving from this wireless system revolution demands a distinct technology aligned to a single infrastructure with the capability to serve users with a high level quality of service across different conditions. Paul et al. (2011) suggest that fulfilling this huge requirement with a wide range of applications need the wireless system of the future generation to be able to communicate across technologies that are heterogeneous such as the WiFi, WiMAX, UMTS, Web 2.0 mobile applications and location-based or car navigational systems [5]. The most desirable feature in the next generation wireless network is its ability to move seamlessly over various access network regardless of the network infrastructure used. The handover between these dissimilar networks can be explored by using vertical handover algorithms. Given the complexity of different access network technologies, a hybrid AI-based architecture provides an effi-

http://dx.doi.org/10.1016/j.neucom.2016.08.136 0925-2312/© 2017 Elsevier B.V. All rights reserved.

Please cite this article as: S. Goudarzi et al., ABC-PSO for vertical handover in heterogeneous wireless networks, Neurocomputing (2017), http://dx.doi.org/10.1016/j.neucom.2016.08.136

JID: NEUCOM 2

ARTICLE IN PRESS

[m5G;March 16, 2017;4:31]

S. Goudarzi et al. / Neurocomputing 000 (2017) 1–19

cient solution with high accuracy, stability and faster convergence for a handover that is vertical in environments that are heterogeneous and wireless. Among the most pertinent features of modern communication is the accessibility to wireless networks using mobile devices, with a high service quality (QoS) based on the preferences of the users. Nevertheless, mobile terminals are able to locate more than a single network of various technologies in its path in the heterogeneous environments, as they have the capacity to connect with other wireless access points based on their values of QoS. It remains that the features of the present mobile devices demand the use of quick and effective algorithms to offer close to realtime solutions as possible. Such limitations have motivated us to create intelligent algorithms that prevent slow and high computing linked to direct search methods thus lowering the time of computation. Since the importance of high latency, packet loss and signaling cost problems during handover process are undeniable, the lack of an effective vertical handover decision (VHD) algorithm, which could select the most optimal access network for handover, is sensible. The complexity of calculating the many parameters in vertical handover decision (VHD) algorithms is another problem. Moreover, it has been shown that the use of adaptive behavior has not been fully investigated. Moreover, a well-established algorithm for a vertical handover decision (VHD) is critically required that would both create a hybrid VHD algorithm which uses forms of intelligence for making decisions via the utilization of mixed heuristic techniques and be able to robustly adapt to the various conditions when the need arises given the dynamic changes that keep occurring in the wireless environment. In terms of the VHO, the components of a network should blend in with the IEEE 802.21 standard of primitives [6] allowing abilities of the MIHF protocol (media-independent handover function) to offer an platform that is homogenous for handovers that are seamless across heterogeneous networks that are wireless, such as the WiMAX, WiFi, LTE, and UMNS. This study presents a vertical handover decision algorithm or otherwise known as VHDA based on two main heuristic algorithms: ‘Artificial Bee Colony (ABC)’ as well as ‘Particle Swarm Optimization (PSO). The proposed VHDA is commissioned by the IEEE 802.21 standard. The remaining parts of this study are organized in the following manner with Section 2 describing the most associated works. The next section describes the network model. Section 4 formulates VHO as optimization problem. Section 5 presents the designed solution. Section 6 discusses on proposed scheme under IEEE 802.21. Section 7 will expound the results of the simulation and Section 8 will present the conclusion. 2. Related works Related works are discussed in three classes as follows: 2.1. Decision-making techniques on the vertical handover process Çalhan et al. (2013) reported a new method using rate of data information, financial costs, and received signal strengths according to various parameters to enable the decision for the handover [7]. The algorithm on the vertical handover decision used within the training procedure is the back propagation algorithm of Levenberg–Marquardt. Values of the weights are modified by this algorithm into a form of group setting, following applications of entire vectors of training. This is among the most successful training algorithms to feed the neural networks forward. Nevertheless, there are several drawbacks to this algorithm. The key drawbacks linked closely to the computing of error functions as well as the Jacobian inversion to get a matrix where dimensions are equivalent

to total weights in the neural network. Thus, the memory requirement is very high [8,9]. This algorithm is also local, and there is no guarantee of finding a global minimum for the objective function (Beliakov and Abraham, 2001). In the case where the algorithm converges to the local minimum, there is no way of escape, and the solution obtained is not optimal [10]. Existing algorithms (ÇAlhan et al., 2013) in [11] take into account the service charges, information on received signal strength or RSSI, and users’ preferences. The algorithm that is proposed as opposed to the conventional algorithm based on the RSSI, significantly improves the outcomes for users and the network because of the proposed fuzzy-based handover techniques. Further, a fuzzybased algorithm greatly lowers the number of handovers in comparison to a SAW-based algorithm. This algorithm is able to switch between GSM, WiFi, UMNS, and WiMAX. Nevertheless, this algorithm has several disadvantages based on the high execution duration that could cause high handover latency. In addition, interface engine inputs could be become more accurate by utilizing AI approaches including the neural network. The research excluded the effect of other environmentally linked determinants and findings in order to examine the mobile parameters of the QoS including the delays in handover as well as the packet loss. He, Qing proposed a study called the VHDA [12] that used the fuzzy logic technique to assess the available networks’ performance. After this, the handover decisions are carried out to choose the most suitable network for the mobile nodes (MNs). This algorithm is applied by considering a combination of various parameters such as received signal strength or RSS, available network bandwidth (B), monetary costs (C), and users’ preference (P) as the vertical handover decision rule. In the existing vertical handover algorithm, each access network is evaluated as a value function. An end user will access the certain network, which maximizes the value function. The VHDA initially gives the vertical handover decision criteria and then the VHDA is employed to calculate the performance evaluation value of the ith network (PEVs) using fuzzy logic theory. Nevertheless, in real conditions, the forms of membership functions must be modified for better impacts. The authors mentioned that the VHDA can make accurate handover decisions, reduce redundant handovers and balance network resources. Also, in the proposed work, the decision process was based on a single fuzzy decision engine, which typically has fixed First Mobile First Service (FMFS) and decision rules, regardless of the traffic types (e.g. voice, video and data). Based on Chandralekha et al. (2010) in [13], the problem of decision is formulated as multiple objective optimization issues while simulating it by the genetic algorithms. The results of simulation demonstrate that the quantity of handovers might be minimized when the values of the optimized network parameter are noted. In addition, this method has lessen the negative impacts of choosing the configuration of the parameter based on the solution quality and the time taken for computing. Jaraíz-Simon et al. in SEFISA [14] developed an adaptive heuristic framework to get optimum network within the Vertical Handover phase of decision, for a terminal mobile, which moves parallel to the heterogeneous wireless networks. Every network is identified based on the QoS parameters values. The problem of optimization tries to adjust the weightage of the QoS to identify the optimal network out of the available ones. They made comparisons of the performances of the SEFISA versus the SEFI in order to formalize the goodness of the adaptive heuristic method. Goodness of the heuristic model was proven to gain optimal solutions, thus, enhancing the performance provided the former exact method and algorithm. Moreover, Xie in [15] VHDA to minimize the cost of grade of service (GoS) in WLAN and cellular networks was suggested. This paper proposed a cost function of the GoS and ob-


JID: NEUCOM

ARTICLE IN PRESS

[m5G;March 16, 2017;4:31]


tained the optimal radius of WLAN by using the simulated annealing (SA) method to minimize cost. Wang et al. [16], and ABC support of the QoS handover decision scheme was suggested, and an optimum solution for handover of the attributing N terminals to the M access networks was discovered according to the well-suited PSO algorithms. Reports based on past literature claim that the access network such as its current load and the terminal such as its current velocity and residual electric capacity, are taken into account, along with the application of the QoS requirements, user preferences across the access network coding system, and user preferences across the access network provider, etc. The optimum Pareto under the Nash equilibrium within the users’ utility as well as the utility by the network provider were reached or reached out to for the said solution with the help of the gaming analysis. Cross layer handover strategies can be projected to offer services that are seamless for mobile terminals within the heterogeneous networks that are wireless. In intending to lower the delay period during handovers, the link layer ought to activate the protocols of handover of the 3 layers in a timely manner as they are able to complete the handover processes prior to the present wireless link terminates. Due to the restricted power of computing within the mobile terminal as well as a bigger rate of packet loss in the vertical handover (Yang et al., 2013) [17], a novel mechanism for triggering based on gray predictions was proposed. First, the duration needed to perform the handover was projected. Second, the time to trigger a Link_Going_Down was identified based on the convex optimization theory, where both took into account the signal strength received from the presently linked network as well as the targeted access network. Simulation’s findings proved the mechanism could achieve more accurate prediction using the similar prediction method [18]; besides, the rate of packet loss could be controlled to 5% where the moving speed of the terminal was 5 m/s or less. In [19], Nadembega et al, proposed a novel dynamic access network selection algorithm which was capable of adapting to prevailing network conditions. Their algorithm was a dual stage estimation process where network selection was performed using sequential Bayesian estimation which relied on dynamic QoS parameters estimated through bootstrap approximation. Simulations demonstrated the effectiveness of the proposed algorithm which outperformed static optimization approaches in a highly efficient manner. However, this algorithm suffers from high computation time. Moreover, according to Ong et al. [20] the network selection problem in heterogeneous wireless networks with incomplete information was formulated as a Bayesian game. Every user has to decide the network selection optimally given only partial information of the preferences of other users. The dynamics of network selection were applied using Bayesian best response dynamics and aggregated best response dynamics. The Bayesian Nash equilibrium was considered the solution of this game, and there was a oneto-one mapping between the Bayesian Nash equilibrium and the equilibrium distribution of the aggregate dynamics. Yetgin et al. [21] introduced a VHO method based on a selfselection decision tree, which could support the VHO among WAVE, WiMAX, and cellular 3 G The decision tree made decisions according to user preferences, and the feedback decision method in line with the feedback of services and movements on vehicles could avoid the negative impact of service changes and movement changes. Also, Chen et al. [22] used a combination of information theory, clustering analysis, and a decision tree algorithm. An energy efficient prediction model was installed on the client. The decision trees allowed a minimum of access points to be used, thus reducing the computation and wake-up time on the client. They developed an algorithm that enabled optimization of the locationestimation accuracy and simultaneously reducing the number of

samples that were needed for high accuracy computation but this method has disadvantage on slow convergence speed. In paper [23], Chakoo et al. proposed an intelligent vertical handover algorithm across heterogeneous wireless networks based on rough set theory. Rough set is used to perform knowledge reduction, taking advantage of the equivalence relation and partitions defined in data sets. Then, they converted the rough set to fuzzyrough set to apply fuzzy decisions. Also, Feng et al. [24] proposed a VHD algorithm based on Fuzzy Logic Decision, a method based on the theory of a rough set forwarded to lower the rules of the FLC as well as chosen central parameters as the criteria for the input. The access network candidacy value was measured utilizing the central parameters in a FLC. Giacomini et al. [25] put forward a new method for the selection of heterogeneous networks that are wireless by implementing the Dempster Shafer Theory, a theory based on mathematical proof. This DST application in two other phases of network selection, namely weighing as well as updating was equally expounded on; these would be able to update the parameter of various needs dynamically as well as the conditions of the based on the various multiple resources. Simulations compared the updated technique of hybrid MDAM verify the suggested DST-based technique’s consistency as well as efficacy. This study extracted some issues faced during the various selections of network phases into a common issue, namely converging various masses (levels of beliefs) gathered based on multiple resources into a combined mass. The application model, implementing and the verifying were also performed in the study. They proved that it was possible to adapt classical data fusion techniques to develop a more robust decision-making mechanism, and moreover that the Dempster–Shafer Theory was the optimal choice. Tamea et al. in [26] highlight on the fact that using conventional QoS parameters does not perform the handover procedure optimally. They have divided this problem into two methods that first method can attain a minimum level of QoS in terms of bit error rate, and second method can reduces the Ping-Pong effect by limiting the number of handovers toward the target APs. In this chapter, we presented some studies on vertical handover decision algorithms based on different techniques and also we reviewed these works based on various methods. In next section, we mainly summarized hybrid methods as promising methods for network selection during vertical handover process in heterogeneous wireless networks.

2.2. Hybrid techniques on the VHO process This section shows the review on different hybrid methods based on used mechanism with other heuristic methods, application domains and so on. Based on literature, the hybrid VHD algorithm utilizes some form of intelligence for decision-making and it is able to robustly adapt to situations regularly due to the necessary dynamic changes in the wireless environment. Lastly, the convergence of the hybrid algorithms to the global optimum solution is better than other intelligence algorithms. There are some related works on hybridization of ABC and PSO on different environments. Several studies have presented results showing that ABC-PSO is better than ABC, such as Shi et al. (2010a) in [27] which proposed a novel hybrid swarm intelligent algorithm based on PSO and ABC. In their method, two information swapping procedures are introduced to share valued information equally between the particle swarm and bee colony. Also, El-Abd (2011) in [28] investigated hybridization of ABC with PSO where the PSO algorithm is augmented with ABC component to improve the personal bests of the particles. Motivated by PSO, Zhu and Kwong [29] implemented an ABC algorithm called gbest-guided ABC, which incorporates


3

JID: NEUCOM 4

ARTICLE IN PRESS

[m5G;March 16, 2017;4:31]


the information of global best solution into the solution search equation. The results of reviewing the most related works are presented in Table 1. This table shows that, as a main result, hybrid algorithms offer better performance because the hybrid algorithm converges more quickly and prevents it from falling into local optima. Also, for better understanding, we illustrated objectives and application domain of each study in this table. 2.3. Protocol for handovers in heterogeneous wireless environments The service protocol, media-independent handover (MIH), has been occupying the IEEE 802.21 Working Group since 2004 and the group’s primary aim is to provide homogeneous function interfaces among the network technologies that are considered heterogeneous. At present, research has been focused on the IEEE 802.21 technology’s performance [41,42]. This standard offers a cooperative usage of the available information at the mobile node as well as inside the network infrastructure; incidentally, the mobile node is in the right position to track the networks that are available. This network infrastructure is suitable for collecting all the information off the network including the mobile nodes’ location, neighborhood cell lists, and the availability of the higher layer service. Decisions regarding connectivity are made by both the network and the mobile node. The major entity of the MIH Function (MIHF) involves a group of functions enabled for handovers inside the network components’ protocol stacks. The fundamental services that the MIHF provides include the media-independent information services or MIIS, media-independent event services or MIES, and the media independent command services (MICS). These services can be interacted by messaging with the lower as well as upper layers. The MIHF services offer services in the year [43]: a) Media Independent Event Services or MIES identifies changes in the properties of the link layer and instigates suitable events (triggers) out of the remote as well as local interfaces; b) Media Independent Information Services or MIIS offers information regarding the various networks in addition to their services which enables further efficient handover decisions to be carried out over the networks that are heterogeneous; and c) Media Independent Command Service (MICS) offers a group of commands for users of MIH to control the link’s properties applicable to the handover in addition to switching between links when necessary. Based on the our knowledge, decision-making algorithms for hybrid vertical handover do not exist within the intelligence schemes, considering the mobile information and the context, the utilization of WiMAX, LTE, and Wi-Fi, as the inherent wireless technologies; the status of the network; preferences of the user; requirements of the functioning application; according to the standard of the IEEE 802.21. 3. Network model Our method of best network selection can be improved by using the standard IEEE 802.21 Media Independent Handover. In fact, the offered method needs information about access networks approximately the MN to make an appropriate decision. Some of the decision inputs for the offered algorithm are acquired over the MIH standard. This protocol assists the progress of the interchange of signaling message between the unit of handover decision and various access technologies. Therefore, MIH is benefited for gaining essential information about network and users. By using the qualifications and the features of this standard, services are achieved without interruption with qualities of service meeting the user’s

requirements. In heterogeneous wireless networks, different networks, including 3GPP (e.g, EDGE, HSPA, UMNS, LTE) and non 3GPP (e.g, WiMax, WiFi) standards, should be interconnected in an optimum way to provide users with a good Quality of Service. This study shows different settings that present the handover signaling for an integrated WiFi, WiMAX, and LTE networks. The first setting demonstrates the signaling whereby an MN is in an overlapped area and can select better connectivity, applying the ABC concept. The MN under the overlapped area of WiMAX and WiFi is depicted in Fig. 1. The second setting explains the signaling for a user obligated to execute the handover as its current connectivity will be lost as it is moving from WiMAX to LTE. All scenarios describe how the MIH framework can offer service continuity to the user’s session as well as some mechanisms involved in this processes. The presented Point of Attachment or PoA and Point of Service or PoS are described in the following in reference to the framework of the MIH. The MIHF in a network entity, which directly communicates with the MIHF in a mobile node, behaves as the PoS of the particular mobile node (MN). The MN exchanges the MIH information with the MIH PoS by utilizing the L3 transport when the PoS is not in residence in the similar network entity as the PoA network. The PoA is the network side of a layer 2 link, which involves the MN as the other end point. Therefore, the framework of the MIH supports the L2 and L3 transports in MIH information exchange. In the handover decision making, two issues should be considered. On one hand, the MN should try make best use of the utilization of a high bandwidth and low cost access network. On the other hand, the numbers of unnecessary handovers have to be reduced to avoid degrading the QoS of current communication and overloading the network with signaling traffic. Each mobile connection may involve a number of vertical handovers during its connection period. The terminal that is mobile is expected to obtain information out of the collocating networks in the range of receiving frequently. Information advertised of every network could possess the bandwidth available as well as the average delay that can be measured using the IETF IP metrics for performance processes. At each time duration, the terminal that is mobile determines if the link ought to utilize present chosen network or be re-directed to some other network that offers a higher performance level (such as reduced costs, better service quality guarantees). The re-directing of the link from one network to another involves a complicated procedure, which improves the network’s load of signaling and processing. Therefore, an exchange occurs between the QoS of the linkage as well as the processing, along with the load of signaling load [20]. 4. Formulate VHO decision making as optimization problem Optimizing the vertical handover process is an important issue in research, as poor optimization leads to a decline in network signaling and mobile device power loss but, on the other hand, advances network quality of service (QoS). In this study, an adaptive heuristic model is designed to achieve the network that is optimal within the Vertical Handover phase of decision, for a terminal that is mobile and moves in random alongside the heterogeneous networks that are wireless. Each network is identified using the Quality of Service parameter values. The problem of optimization tries to align the Quality of Service weights to establish the network that is optimum among the networks that are available. This study has demonstrated the goodness of the heuristic model in achieving the solutions that are optimal, enhancing performance provided by the prior exact method as well as other algorithms [44]. The efficient adjusting capability of the QoS weights to establish the network that is the best out of the networks that are available


JID: NEUCOM

Objective

Method/s compared with

Application domain

Result

Literature

ABC, PSO

To enhance the efficiency of energy consumption, to deploy sensor nodes at optimal locations and to schedule these sensor nodes. To maximize QoS and networks aim to maximize their profit, which can lead to improve the individual efficiency of mobile users. To reduce system load.

PSO

Wireless sensor networks.

Artificial bee colony algorithm performs better for sensor deployment problem.

[30]

Greedy algorithm.

Heterogeneous wireless networks.

Better performance.

[31]

Traditional fuzzy logic algorithms.

Heterogeneous networks. Heterogeneous networks. Heterogeneous networks. Heterogeneous networks.

wireless

Better distinguish ability

[32]

wireless

Better performance where prediction accuracy is greater. Lowers computational difficulty greatly, and assures the HO decision performance. PSO-FNN can balance the load of heterogeneous wireless networks effectively. Better performance than that of other bandwidth reservation schemes in terms of CDP and CBP. High performance in heterogeneous as well as homogeneous network environments. A rather appropriate algorithm for precision as well as computational time to overcome the problem of optimization. Proposed method has better performance.

[33]

GA, Game theory

DE, FL GRA, PSO, FL RST, FL

To minimize the number of handovers, reduce calculation time. To reduce the computational complexity.

PSO-FNN

To decrease the blocking probability as well as handover call blocking probability.

GRA, PSO

To reserve bandwidth for the handovers, decrease the call blocking probability.

FL, GRA

To reduce the call dropping probability.

SA, GA

To decrease computing time.

NN, FL

To guarantee network connectivity, reduced call drop rate. To minimize energy consumed and end-to-end delay. To enhance the accuracy of malicious detection. To effectively reduce ranging error and location error. To reduce the number of unnecessary HOs and to minimize the HO delay. To improve the individual efficiency of mobile users.

DE, GA Fl, ICA PSO, Chaos, NN. NN-FL NN-FL

Self-tuning algorithm. Traditional FLC-based VHD algorithm. Sum-received signal strength (S-RSS) algorithm. Fixed reservation scheme (FR) and rate-based borrowing scheme (RBB). The quantitative decision values of the candidate networks. AHP and FMADM.

wireless wireless

Heterogeneous wireless networks. Heterogeneous wireless networks. Heterogeneous wireless networks.

Comparison with classical, fuzzy approaches. Traditional HO algorithms.

Heterogeneous wireless networks. Wireless ad hoc networks.



Kinds of location algorithms.


Traditional HO algorithms.

Heterogeneous networks.


Heterogeneous cognitive networks.

[24] [34]

[35]

[36] [37]

[38]

Better performance.

[21]

Performs better than traditional fuzzy logic algorithms. Proposed method has better performance.

[39]

Performs better than traditional algorithms for handover time in hybrid networks. Better performance.

ARTICLE IN PRESS

Hybrid


[40] [39] [40]

5

[m5G;March 16, 2017;4:31]

Please cite this article as: S. Goudarzi et al., ABC-PSO for vertical handover in heterogeneous wireless networks, Neurocomputing (2017),

http://dx.doi.org/10.1016/j.neucom.2016.08.136

Table 1 Hybrid methods.

JID: NEUCOM 6

ARTICLE IN PRESS

[m5G;March 16, 2017;4:31]


Fig. 1. Heterogeneous wireless networks.

is critical in the networks that are wireless. We need to have knowledge of the merit of each available network to identify the network that is the best. To achieve this, we are required to design a function or metric that would acquire this network’s merit. First, a set of weights assigned to each one of the QoS parameters is used to measure the quality of the network, and they are based on the network characteristics and user’s preferences. A general profile including any quality of service parameter can be any weight from 0 to 1 assigned to it. This measurement is offered by a particular function, known as the cost function. The function is assessed in the VH stage of decision-making. Hence, the problem of optimization consists of searching for the best solution that, when applied to each network, returns the lowest cost to the network, which will then be selected as the solution that is best for the VH stage of decision. The Artificial Bee Colony or the ABC component assigns a suitable weight (w1 , w2 , …, wi ) for each initial decision according to the objective function that is specified by the operator according to the importance and sensitivities of the access network selection criteria to the different characteristics of a wireless heterogeneous environment [45]. We have developed some novel algorithmic proposals deep in the intelligent computing that solve the optimization problem, which is to find the best combination of weights for the quality of service parameters of heterogeneous wireless networks for a mobile terminal. 5. Solution design At this stage, a decision-making algorithm for hybrid vertical handover that can choose the best network that is applicable is proposed. The scheme, which is proposed, comprises two major heuristic algorithms namely the ‘Artificial Bee Colony or ABC’ [46,47] as well as the ‘Particle Swarm Optimization or PSO’ [48]. Algorithms that are metaheuristic algorithms utilize various exploitation and exploration techniques to address the issues of high-dimensional optimization. With the objective of overcoming the weak exploration capability of the PSO and the weak ABC mechanism for exploitation, the algorithm comprising the hybrid metaheuristic is a latest research approach in overcoming the issues of high-dimensional optimization that have garnered reasonable interest in the past few years. The algorithm for hy-

brid metaheuristic in this study is the combined process for the hybridization of the PSO and the ABC. The ABC and PSO hybridization selected for the process of decision-making is the ABC-PSO algorithm that applies the ABCPSO on the entire networks that are available and accessible at present. A new hybrid process (ABC-PSO) is proposed to hybrid the PSO and the ABC by utilizing the exploitation capability of the PSO in addition to the ABC’s exploration capability. In general, 3 types of bees go about looking for the food source or solutions in the ABC such as the employed bees, the onlooker bees, as well as the scout bees. Employed bees search for food close to the source of food based on their memory, and after this, they transfer this information on food to the onlooker bees. Onlooker bees will then choose the better sources of food amongst the ones discovered by the employed bees and simultaneously looking for more food close to the prior chosen areas of sources of food. Several scouts are then engaged amongst the employed bees that will leave their previous sources of food and look for new sources. According to (Karaboga and Basturk, 2007), the ABC algorithm framework comprises several essential components. First, the process of update utilized in the onlooker phase is similar to the phase of the employed bees. Moreover, a greedy choosing is carried out in the process of update as well as the onlooker phases. Second, at the onlooker phase, the selection of solutions is done based on the rule of probability [46]. Third, the primary differentiation between the phases of employed and the onlooker bees is that in each solution at the phases of employed bees phase, the process of update is carried out while at the onlooker phase, only the chosen solutions have the chance of being updated. Fourth, at the scout phase, a solution that is inactive relates to a solution that remains unchanged over several generations [47]. Eberhart and Kennedy (1995) who developed the particle swarm optimization (PSO) [48] were motivated by the social behaviors of a group of birds attempting to reach a destination that is not yet known. All the solutions are reflected using one ‘bird’ from the flock as well as this is symbolized using a particle. This particle is exactly like the food source within the ABC. In comparison to the ABC, the PSO represents a source of food within the ABC. In comparing with the ABC, the PSO does not give rise to new ‘birds’ in the evolution procedure. Nevertheless, the birds found in


JID: NEUCOM

ARTICLE IN PRESS

[m5G;March 16, 2017;4:31]


the population develop behaviors and after that go to the destination based on trajectories of prior movements. In reality, these birds interact as they are flying. Every bird flies in a particular direction. In the midst of the flight, first, a bird pinpoints the best bird’s position (global experience. At the same time, it also examines the best positions of the historical records (local experiences). Lastly, the bird’s new direction is fixed by combining local as well as global experiences. These birds attempt to identify the best optimal position. Just like the ABC, this process is repeated until the destination that is desired is located. It should be noted that all the birds learn both from their own experiences and from the group’s local experiences. Firstly, the ABC develops N solutions (particles/honeybees) consisting of randomly distributed initial population, whereby N represents the quantity of the bees, either employed or onlookers. Next, the velocity, g-best, and the p-best of every bee are provided with initial values similar to the PSO’s initial process. Every solution of Xn (n = 1, 2,…, N) is a vector with D-dimension, whereby d represents the quantity of the optimization parameters. The positions’ population (solutions) goes through a repetition of cycles as in C = 1, 2, maximum number of cycles (MCN) by the employed bees, onlooker bees, as well as the scout bees’ search processes. The function in the following examines the source of food created by the fitness function, whereby the fitness function’s design ascertains what should be optimized by the ABC. Each food source’s cost can be ascertained using the cost function that is intended to be optimized. The costs function (costi ) in this study is designed in the following manner in Eq. (1),

1 costi = CN + 1

(1)

whereby CN denotes the ith solution’s objective value. In continue, we show the details of calculating cost function. Clearly, the network choice that results in the lowest calculated value of the cost function is the network that would provide the most benefits to the user. In terms of network conditions, network related parameters such as available bandwidth and network latency may need to be considered for effective network usage. Use of network information in the choice to handover can also be useful for load balancing across different networks. In addition, a variety of parameters can be employed in the handover decision to guarantee the system performance, for example, the bit error rate (BER). Mobile terminal conditions also include dynamic factors such as velocity, moving pattern, moving histories, and location information. For any of the service types, two sets of relative priorities are defined. These two sets are interface priorities and application priorities where relative priorities among available interfaces in a device are interface priorities (Ahmed et al., 2006) [49] and relative priorities among five types of services are application priorities (Goudarzi et al., 2016) [50]. At this stage, suitable limit values (upper and lower) for the five QoS parameters are mapped at the backend for each of the five service types. While fixing the limit values, it is important to note that high values are not always better for all the five QoS parameters. It is always preferable to have values as high as possible for bandwidth, but as low as possible for delay and BER. These values are based on the contexts such as QoS requirements of specific service types and interface capabilities. At the next stage, ranking of the available networks is performed based on the interface priorities scores and application priorities scores assigned at stage 1. Consider a set of candidate networks S = {s1 ,…, sN } and a set of quality of service factors Q = {q1 ,…, qM } where N is the number of candidate networks and M is the number of quality of service factors. In addition, we consider that each QoS factor has its own weight and this weight shows the effect of the factor on the network or user. Thus, cost

7

function for each network can be calculated using Eq. (2) where W N is calculated by the Analytic Hierarchy Process (AHP) (Saaty, 1988) [51]. This process is chosen due to its ability to vary its weighting between each factor based on network conditions and user preferences.

CN = Winter f ace ×

M

q j × Wj

(2)

j=1

With the above definitions, the AHP method is described as follows: Subsequently, the relative scores among the QoS scores set are calculated. Relative scores between any two particular scores are calculated using Eq. (3) where Rqi q j is the relative score between parameters qi and qj , and Sqi and Sqj are their respective scores.

⎧ Sqi ⎪ ⎪ Rqi q j = 1 − × 10; ⎪ ⎪ Sq j ⎨ 1

Rq j qi = ; ⎪ ⎪ Rqi q j ⎪ ⎪ ⎩

j>i (3)

j 0) then if (Dist_Recep_Prob (location, destination, MIIS_pkt) > 0) then a ← true Initialize velocity, p-best and g-best return (a) while iter < Max MCN do for i = 1 ..N do FOR (Updating velocity for initial networks){ Update velocity and position for MN Update_current_nghbhood (loc,dist); Evaluate C Fvalue Filter networks via particle filters 2 Step 2: The employed bee phase Procedure CHECK_NGBHDING(time,prdc_wdw,snsg_prd) Evaluate C Fvalue Filter networks via particle filters Update velocity and position for MN 3 Step 3: The onlooker bee phase Procedure CHECK_NGBHDING(time,prdc_wdw,snsg_prd) for i = 1 ..N do Calculate probability for each C Fvalue via Gaussian particle filtering (GPT) Generation of new networks in the neighborhood of onlooker bees Evaluate C Fvalue Filter networks via particle filters Obtain an optimal solution Check location(location, time) Check destination(destination, time) Update app_ preferences (MIIS_pkt,weights_factor_list) Update user_ preferences (list of user preferences) 4 Step 4: For (scout’s bee phase){ if (check_neighbor(location, destination)> 0)then Evaluate C Fvalue Filter networks via particle filters if C Fvalue ≥ limit then Update velocity and position for MN Update (location, destination) for current networks; Update velocity and position for MN Schedule (available networks (next time)) Select best network based (global optimal solution) a ← CFvalue (app_req,user_ pref, QoS _req) Memorize the best network at each cycle return (void) handover(a, event); return (void) if the stop criterion is satisfied then go to end while else return to Step 2 End while Procedure SENSING_INTERFACES Procedure SENSING_INTERFACES (void) repeat event ← query of exiting interfaces(location, destination, exiting interfaces_MICS_packet, MIES_packet); if event = Link Detected or Link going Down then {go to Decision_Making Phase (event); return (void) Procedure Cost _function Procedure Cost _function (void) repeat Update application request list (MIIS_weights_factor_list) Update list of user preferences Update QoS_req ← QoS_req(MIIS_weights_ QoS _list) CFvalue ← CFvalue (app_req,user_ pref, QoS _req) return (void)

After the MN determines the HO to a network that is new, because of loss of signal quality or because of users’ preference, the MN activates the mechanism MIH_Get_Information Request to gain information simultaneously regarding the CRANs around the MN as well as the resources available in every one of them (process


ARTICLE IN PRESS

JID: NEUCOM


[m5G;March 16, 2017;4:31] 11

Fig. 3. MIH framework combined with vertical handover decision algorithm.

Fig. 4. MN triggers to retrieve information about the CRANs.

of CRANs Discovery as well as Resources Availability Check). The mechanism proposed in the study, which is the IS, has an updated information that is real-time regarding the resources of the radio that are available in every underlying RAN. In algorithm (ABC-PSO), some RATs PoAs directly transmit the MIH_Context_Information Indication message to the algorithm of the ABC-PSO. As a result, the IS is able to update its database with the new resource information. The signaling process is presented in Fig. 4. As demonstrated in Fig. 5 in detail, some of the RATs PoAs directly transmit the MIH_Context_Information Indication message to the IS. As a result, the IS is able to update its database with the new resource information. After the MIH_Get_Information Response message is received and processed, the mobile node initiates the scanning procedures

for every radio access network obtained in the previous step. After the step of scanning, the mobile node filter networks through a particle filters or (Sequential Monte Carlo Methods). Continuing, the resources for every radio access networks are compared with the threshold of the Data rate. The MIH_LINK_GOING DOWN event of the MIH standard is according to the RSS from the present network. The data rate level information from the PHY layer is the first priority parameter to establish the right time for initializing the handover. After a WiMAX or Wi-Fi interface is found and the data rate is acceptable, the terminal will activate the process of network selection. The structure of the current event MIH_LINK_GOING DOWN is modified using the data rate to replace the RSS. The data rates (θ ) required by the different applications are monitored by the MN


ARTICLE IN PRESS

JID: NEUCOM 12

[m5G;March 16, 2017;4:31]


Fig. 5. CRAN discovery.

Fig. 6. Filtering access networks and updates MN’s velocity.

regularly; a handover process is started by the MN when the θ goes lower than the threshold that is pre-defined. For example, the value of a threshold that is pre-defined would be 256 kbps when the minimum needed data rate via a streaming app. is 256 kbps. In addition, the handover will be initiated by the MN, when the MN runs a streaming application and checks the data rate requirement; if it is lower than 256, then the handover will be initiated by the MN. The signaling process is presented in Fig. 6. The movement of the MN with variable speeds in three various networks were considered. A few experiments were performed with various MN speeds. The velocity of the MN in the proposed algorithm can be updated in two various phases to increase the network selection accuracy as the MN has movement and the Dwell time measurement depends on the speed and moving pattern of the user as a metric for selection. The cost function of every radio access network should be established (Line# 10) based on the MN’s new position. The proposed algorithm updates the database with new resources information for this aim. This is done to reach viewpoint that is global of the entire networks in the location that are heterogeneous to

optimize the VHOs that is seamless while roaming is carried out over the networks. Then, the radio access networks are filtered utilizing the particle filters or (Sequential Monte Carlo Methods). The Gaussian particle filtering or GPF is utilized for measuring the onlooker bees’ probability to the food sources rather than the method of the roulette wheel selection. This is approach is considered the least complex compared to the other methods. This method of implementation is simple. An algorithm based on the GPF does not need a sampling operation and therefore it is easy to amend it for implementing on a parallel basis. Following this step, it is possible to obtain an optimal solution (Line #15). The signaling process is presented in Fig. 7. After the onlooker bee phase, if an unfit network is present for selecting relying on the “limit”, then the network that is unfit is replaced with a new randomly network for the scout bees phase (Line #19). The signaling process is presented in Fig. 8. The proposed ABC-PSO algorithm can be applied on vehicular networks (VNs). This property is due to the fact that, the heterogeneity of such technologies, rather than being a pitfall for vehic-


JID: NEUCOM

ARTICLE IN PRESS S. Goudarzi et al. / Neurocomputing 000 (2017) 1–19

[m5G;March 16, 2017;4:31] 13

Fig. 7. Obtain an optimal solution.

ular communications, should be seen as an advantage, since vehicles can make the most out of the diverse wireless technologies to maintain continuous communication while journeying from one location to another [56]. Based on MIH’s operation of preparation scenario, immediately after the handover is activated, the MN transmits a request for information to the IS located in the core network, via its present serving network (SN). The MN in the information request could contain information regarding its subscription profile, user preference, present location, velocity, power status, and the MN’s demands for running the applications. Therefore, the IS can establish a sorted list of networks that are available based on the MN needs, according to the available static information regarding the supported rate of data of every network technology, pricing, coverage, energy consumption, etc. After the MN gets the information reaction along with the sorted candidate networks list, the particular area of the band is scanned and the resources (QR) of the first candidate are queried.

The VHO is executed if the reaction is positive. If the first candidate network is unavailable, the SN transmits again a ‘query resources’ message, to the following candidate, until a selection of a target network is successful or until the candidate networks from the list are exhausted. The major entities in the scenario mentioned above are several NRANS of RANs that serve the area being studied, several NMN of MNs in every RAN, and the IS. It should be noted that every RAN behaves both as the SN of its own MNs and as a candidate network (CN) for (at least several of) the MNs in other SNs. In addition, the exchanges of the signaling message that will be explained hereafter happen across communication links between the abovementioned entities. Communication between a MN and its SN in particular, happens across the wireless link offered by the corresponding RAN. Communication among the various RANs and between a RAN and the IS happen across wired links of the network’s backbone. Continuing, the handover delay, packet loss, and cost of signaling


ARTICLE IN PRESS

JID: NEUCOM 14

[m5G;March 16, 2017;4:31]


Fig. 8. Selection best network.

Table 2 ABC-PSO elements. Elements

WiFi

WiMAX

LTE

Access points Data rates (Mbps) VHO latency (ms) Coverage(m) Advertisement interval

2 28.5 1080 500 100

1 17.3 2665 10 0 0 50 0 0

1 3.5 – 50 0 0 –

for the MIH-based vertical handover signaling process that is proposed while the handover occurs between the various networks are analyzed. 7. Results and discussion The NS2 is utilized to assess the network performance. Tables 2 and 3 denote the ABC-PSO algorithm parameters. The average time between the successive decision epochs is determined at 15 s. The unit of bandwidth is determined at 16 kb/s, the jitter unit is determined at 2.5 ms, and the traffic unit is determined at 0.5 erl. The minimum and maximum velocities are determined at 1 unit and 5 units, respectively as shown in [57,58]. The cellular area is three times that of the WLAN, and the MNs’ spatial density in the cellular network is eight times bigger than that found in the WLAN. The WiMax DL’s peak data rates are 75 Mbps UL: 25 Mbps, and in the LTE DL: 100 to 324.6 Mbps UL: 50 to 86.4 Mbps. Our proposed VHO algorithm is evaluated with the Technique for Order of Preference by Similarity to Ideal Solution or otherwise known as TOPSIS (Technique for Order Preference by Similarity to the Ideal Solution)

Fig. 9. Average numbers of HOs under different MN’s velocities (c = $ 90 dBm).

[59] according to the average quantity of handoffs, the available bandwidths, etc. Figs. 9–17 denote the performance of the network in the handoff scenario. The MIIS offers information regarding the networks that are available and their respective PoAs contained in the simulated space. The main configuration set for the experiments is summarized in Table 2. One LTE, two Wi-Fis, and one WiMAX PoA are observed covering the various areas with distinct data rates offered. In addition, every network in this situation is configured using dif-


ARTICLE IN PRESS

JID: NEUCOM

[m5G;March 16, 2017;4:31]


15

Table 3 Summary of network parameters. PoA

Technology

Price (MB)

Latency (Packet)

Ratio of packet loss

Throughput (Mbps)

PoA-1 PoA-2 PoA-3 PoA-4

WiFi WiMAX LTE Wi-Fi

0.8 0.15 0.9 0.075

15.44 17.59 25.22 35.08

1.19 2.74 0.76 3.1510

1.48 1.18 1.42 0.59

Table 4 Weights values of cost function for user preferences. AP

Cost

Streaming

Conversational

Maximum performance

Latency Packet loss Throughput Price (MB)

0.1057680 0.0224017 0.3733550 0.4534210

0.4345300 0.1432380 0.0234577 0.4176509

0.1657905 0.2545390 0.3435154 0.2467543

0.0623170 0.3898690 0.4845670 0.0367890

Fig. 11. Average number of HO versus average velocity. Fig. 10. Decision delay versus average velocity.

ferent parameters of performance. Different alternatives are developed to assess the CNs using this method. Table 3 shows the set of parameters of all the networks while Table 4 demonstrates the minimum requirement for video sessions to be reached using the selected network in the simulation. As the video function is projected to be a key function in the increment in future mobile application demand, we have focused on video streaming traffic. The average time between successive decision periods is set to 15 s. The bandwidth unit is set to 16 kb/s, the jitter unit is set to 2.5 ms, and the traffic unit is set to 0.5 erl. The maximum and minimum velocities are set to 5 units and 1 unit, respectively as in [60,61]. The area of the cellular is as three times as that of WLAN, while the special density of MNs in the cellular network is eight times larger than that in the WLAN. As displayed in Fig. 9, we can detect that the ABC-PSO algorithm reduces unnecessary number of HOs by considering the relationship between the change of the MN’s speed and the handover threshold. We also detect that the number of HOs decreases as the velocity increases due to the decline of the connection period in the current network. From Fig. 10, we can find that as the value of average velocity drops, the decision delay increases. This is because the MN’s velocity can be updated in different phase which are affected by the average velocity and threshold values, by which the MN’s connection time is increased.

Fig. 12. Available bandwidth under different MN’s velocities (c = $ 90 dBm).

Fig. 11 demonstrates the number of HOs under different average velocity values. It can be observed that this number decreases as the value of average velocity reduces. For the reason that the decline of average velocity extends the connection time then can reduce the possibility of HO. It can be found in Fig. 12 that our offered algorithm modifies the handover thresholds dynamically according to the velocity of


JID: NEUCOM 16

ARTICLE IN PRESS

[m5G;March 16, 2017;4:31]


Fig. 13. Available bandwidth versus average velocity. Fig. 15. MN’s total cost under different average connection durations.

Fig. 14. Total cost versus delay threshold (v = 2.6 m/s).

MN. We can also observe that the bandwidth decrease as the velocity increases. It should be mentioned that the simulation results presented in Fig. 13 are obtained under the same SST value (c = $90 dBm), which we may modify to satisfy the requirements of different MNs. In Fig. 13, we can see that the available bandwidth of MN decreases as average velocity increases. Because when the value of average velocity decreases, MN can stay longer in network. Therefore, the network can consume network resource efficiently via adjusting the velocity of MN. Fig. 14 shows the relationship between delay and total cost. We can observe that when the delay threshold decreases, the total cost increases under the same SST value. Therefore, we can adjust the SST value to decrease the network cost, by which we can reduce network consumption and guarantee MN’s QoS. We then compare the proposed algorithm with TOPSIS scheme in average number of HOs, etc., and Figs. 15–17 demonstrate the network performance in multi-attribute handover situation. In the case of constant bit rate (CBR) voice traffic, the expected total cost under different average connection durations is shown in Fig. 15. We can observe that our algorithm achieves the lowest expected total cost, as it takes the MN’s different requirements into account and considers the velocity during the vertical handover procedure by merging AHP method to control the importance of

Fig. 16. Comparisons of MN’s total cost under different velocities.

the QoS factors. Consequently, our proposed algorithm outperforms TOPSIS scheme. Fig. 16 shows MN’s total cost under different velocities. We can find that, the total cost of our algorithm improves as the velocity increases, and keeps steady within a certain range of velocity, while the network performances in TOPSIS algorithm decline considerably. This is because TOPSIS has no consideration for MN’s velocity, which can be leads to the broken connection during the handover process. Additionally, QoS degradation, more vertical handover signaling and processing costs also increases the cost. Fig. 17 clarifies the simulation results of total cost under different handover signaling loads. The total cost decreases as the handover signaling load increases, because signaling load rises during each connection which results in the decline of actual cost. Our proposed algorithm lowers the probability of call dropping, together with the cost of signaling and processing, by taking MN’s velocity into consideration. Therefore, the reduction of the total cost is less than that in another algorithm. It can be found that, as the handover signaling load increases, the average number of handovers decreases. The increasing handover signaling load makes the actual total cost of the candidate network lower than the cur-


JID: NEUCOM


[m5G;March 16, 2017;4:31] 17

cisions will be adopted under our method where a mobile terminal realizes different types of wireless access networks with their corresponding QoS values and constantly applies ABC-PSO to evaluate performance and behaviors, among many other possibilities. References

Fig. 17. Comparisons of total reward under different handover signaling loads.

rent one in which the MN resides. Therefore, the proposed algorithm can avoid a number of unnecessary handovers. In terms of cost, we can detect that the various user profiles are also related to various costs. We can find that the minimum cost profile was able to reach the original goal by selecting the networks accurately. Moreover, we can observe that the maximum performance profile achieves the maximum performance such as high throughput, low latency, and low packet loss rate through choosing the candidate networks with better-quality performance but paying the maximum cost for those high-quality services. 8. Conclusions Future wireless communication system will contain various kinds of wireless access networks, and seamless vertical handover between different networks is a challenging problem in HWNs. Although many vertical handover decision algorithms have been proposed, most of them do not consider the impact of MN’s velocity during the vertical handover decision process. In addition, most of current multi-attribute vertical handover algorithms cannot predict MNs’ circumstances dynamically. In order to guarantee the QoS of different MNs, we have proposed ABC-PSO vertical handover decision algorithm for single and multi-attribute situations, respectively, with the objectives of minimizing the expected total cost and minimizing the average number of handovers. Our work has considered the MN’s velocity, network access cost and switching cost of the vertical handover decision, and constructed a cost function to model the QoS properties. Furthermore, we have applied the AHP method to calculate the weight of each QoS factor in the cost function, and adopted the proposed algorithm under MIH to provide link intelligence and other related network information to upper layers. By taking the MN’s velocity into consideration, our work can avoid unnecessary handovers. We have made comparisons to the TOPSIS algorithm to assess the performance of the network. Numerical results demonstrate that the ABC-PSO algorithm has superiority in the number of handover, available bandwidth, and decision delay. The proposed algorithm has a lower total cost and lower average number of handover than the TOPSIS algorithm. Finally, we want to consist VANET technologies as a future research plan. On the one hand, we plan to improve ABC-PSO optimized code for infrastructure-based VNs rather than VANET-based solutions .On the other hand, we want to use protocols (e.g., DSRC and IEEE 802.11p) which capable to get the information collected through car-to-car communications and to deliver such information to the MIIS databases to enhance the knowledge and the de-

[1] K. Gai, M. Qiu, H. Zhao, L. Tao, Z. Zong, Dynamic energy-aware cloudlet-based mobile cloud computing model for green computing, J. Netw. Comput. Appl. 59 (2016) 46–54. [2] X. Liu, C.W. Yuan, Y. Li, Z. Yang, B. Cao, A Lightweight algorithm for collaborative task execution in mobile cloud computing, Wireless Personal Commun. 86 (2) (2016) 579–599 2016. [3] M. Sookhak, A. Gani, M. Khurram Khan, R. Buyya, Dynamic remote data auditing for securing data storage in cloud computing, Inf. Sci 380 (2015) 101–116. [4] Huang, F., C. Wen, H. Luo, M. Cheng, C. Wang, and J. Li. Local quality assessment of point clouds for indoor mobile mapping. Neurocomputing, vol 196, pp. 59–69. [5] S. Paul, J. Pan, R. Jain, Architectures for the future networks and the next generation internet: a survey, Comput. Commun. 34 (1) (2011) 242. [6] IEEE Standard for Local and Metropolitan Area Networks, Media Independent Handover Services, IEEE P802.21-2008, 2009. [7] A. Çalhan, C. Çeken, Artificial neural network based vertical handoff algorithm for reducing handoff latency, Wireless Personal Commun. 71 (4) (2013) 2399–2415. [8] Qing He, A fuzzy logic based vertical handoff decision algorithm between WWAN and WLAN, in: 2nd International Conference on Networking and Digital Society (ICNDS), 2010, Vol. 2, IEEE, 2010. [9] CP Behera, PK Behera, Minimization of number of handoff using genetic algorithm in heterogeneous wireless networks, Int. J. Latest Trends Comput. 1 (2) (2010) 24–28. [10] G. Beliakov, A. Abraham, Global optimisation of neural networks using a deterministic hybrid approach, in: Proceeding of 1st International Workshop Hybrid Intelligence System Hybrid Information System, Adelaide, Australia, Dec. 11–12, 2001, pp. 79–92. [11] A. ÇAlhan, C. ÇEken, Case study on handoff strategies for wireless overlay networks, Comput. Standards Interfaces 35 (1) (2013) 170–178. [12] Qing He, A fuzzy logic based vertical handoff decision algorithm between WWAN and WLAN, in: 2nd International Conference on Networking and Digital Society (ICNDS), 2010, Vol. 2, IEEE, 2010. [13] CP Behera, PK Behera, Minimization of Number of Handoff Using Genetic Algorithm in Heterogeneous Wireless Networks, Int. J. Latest Trends Comput. 1 (2) (2010) 24–28. [14] M.D. Jaraíz-Simon, et al., Simulated annealing for real-time vertical-handoff in wireless networks, Advances in Computational Intelligence, Springer, Berlin, Heidelberg, 2013, pp. 198–209. [15] S. Xie, Vertical handoff decision algorithm based on optimal grade of service, IETE J. Res. 56 (1) (2010) 44–51. [16] X-W Wang, et al., Niche PSO based QoS handoff decision scheme with ABC supported, in: IEEE International Conference on Intelligent Computing and Intelligent Systems, 2009. ICIS 2009, Vol. 3, IEEE, 2009. [17] J. Yang, X. Ji, A Vertical Handover Trigger Mechanism Based on Gray Prediction, J. Theoret. Appl. Inf. Technol. 48 (2) (2013). [18] S. Goudarzi, W. Haslina Hassan, A.-H. Abdalla Hashim, S.A. Soleymani, M.H. Anisi, O.M. Zakaria, A novel RSSI prediction using imperialist competition algorithm (ICA), radial basis function (RBF) and firefly algorithm (FFA) in wireless networks, PLoS One 11 (7) (2016) e0151355, doi:10.1371/journal.pone. 0151355. [19] A. Nadembega, A. Hafid, T. Taleb, A destination & mobility path prediction scheme for mobile networks, IEEE Trans. Veh. Technol. 64 (6) (2015) 2577–2590. [20] EH Ong, JY. Khan, Dynamic access network selection with QoS parameters estimation: a step closer to ABC, in: Vehicular Technology Conference, 2008, VTC Spring 2008. IEEE, 2008. [21] H. Yetgin, KTK Cheung, L Hanzo, Multi-objective routing optimization using evolutionary algorithms, in: Wireless Communications and Networking Conference (WCNC), 2012 IEEE, IEEE, 2012. [22] Y. Chen, et al., Power-efficient access-point selection for indoor location estimation, IEEE Trans. Knowl. Data Eng. 18 (7) (2006) 877–888. [23] N. Chakoo, S. Bahk, Rough Set Approximation Framework for Smarter Vertical Handovers, in: Proceedings of Symposium of the Korean Institute of communications and Information Sciences, 2010, pp. 202–203. [24] Y. Feng, et al., A low-complexity fuzzy-logic-control-based vertical handoff decision algorithm with rough set theory, in: 2013 International Conference on Wireless Communications & Signal Processing (WCSP), IEEE, 2013. [25] D. Giacomini, A. Agarwal, Vertical handover decision making using QoS reputation and GM (1, 1) prediction, in: 2012 IEEE International Conference on Communications (ICC), IEEE, 2012. [26] G. Tamea, M. Biagi, R. Cusani, Soft multi-criteria decision algorithm for vertical handover in heterogeneous networks, IEEE Commun. Lett 15 (11) (2011) 1215–1217. [27] X. Shi, Y. Li, H. Li, R. Guan, L. Wang, Y. Liang, An integrated algorithm based on artificial bee colony and particle swarm optimization, in: 2010 Sixth International Conference on Natural Computation, 2010, August.


JID: NEUCOM 18

ARTICLE IN PRESS

[m5G;March 16, 2017;4:31]


[28] M. El-Abd, A hybrid ABC-SPSO algorithm for continuous function optimization, in: 2011 IEEE Symposium on Swarm Intelligence (SIS), IEEE, 2011, April, pp. 1–6. [29] G. Zhu, S. Kwong, Gbest-guided artificial bee colony algorithm for numerical function optimization, Appl. Math. Comput. 217 (7) (2010) 3166–3173. [30] S. Mini, S.K. Udgata, S.L. Sabat, Sensor deployment and scheduling for target coverage problem in wireless sensor networks, IEEE Sensors J. 14 (3) (2014) 636–644. [31] C. Zhang, X. Wang, M. Huang, A multi-objective genetic algorithm based handoff decision scheme with ABC supported, Intelligent Computing Theories, Springer, Berlin Heidelberg, 2013, pp. 217–226. [32] L. Xia, LG. Jiang, C. He, A novel fuzzy logic vertical handoff algorithm with aid of differential prediction and predecision method, in: Proceeding of IEEE ICC’06, 2006, pp. 5665–5670. [33] S. Venkatachalaiah, R.J. Harris, J. Murphy, Improving handoff in wireless networks using Grey and particle swarm optimisation, in: CCCT, vol. 5, 2004, pp. 368–373. [34] Wang Nan, et al., PSO-FNN-based vertical handoff decision algorithm in heterogeneous wireless networks, Proc. Environ. Sci. 11 (2011) 55–62. [35] C-J Huang, H.-Y Shen, Y-T Chuang, An adaptive bandwidth reservation scheme for 4 G cellular networks using flexible 2-tier cell structure, Expert Syst. Appl. 37 (9) (2010) 6414–6420. [36] X Liu, L-g Jiang, A novel vertical handoff algorithm based on fuzzy logic in aid of grey prediction theory in wireless heterogeneous networks, J. Shanghai Jiaotong Univ. (Sci.) 17 (2012) 25–30. [37] M.D. Jaraiz-Simon, J.A. Gomez-Pulido, MA. Vega-Rodriguez, Embedded intelligence for fast QoS-based vertical handoff in heterogeneous wireless access networks, Pervasive Mobile Comput. 19 (2014) 141–155. [38] A. Singhrova, N. Prakash, Vertical handoff decision algorithm for improved quality of service in heterogeneous wireless networks, IET Commun. 6 (2) (2012) 211–223. [39] S. Shamshirband, et al., D-FICCA: a density-based fuzzy imperialist competitive clustering algorithm for intrusion detection in wireless sensor networks, Measurement 55 (2014) 212–226. [40] H. Wang, et al., A cross-layer bandwidth request algorithm based on chaotic prediction model for broadband multimedia satellite communications, in: National Doctoral Academic Forum on Information and Communications Technology 2013, IET, 2013. [41] A. De La Oliva, L. Eznarriaga, C.J. Bernardos, P. Serrano, A. Vidal, IEEE 802.21: a shift in the media independence, in: Proc. Future Netw. Mobile Summit, Jun. 2011, pp. 1–8. [42] K. Inoyatov, S. Tursunova, I. Kim, Y.-T. Kim, IEEE 802.21 MIH in Linux Kernel space for cognitive and smart handovers, in: Proceeding of IFIP/IEEE International Symposium IM, May 2011, pp. 674–677. http://dx.doi.org/10.1109/inm. 2011.5990653. [43] IEEE P802.21/D11.0 Draft Standard for Local and Metropolitan Area Networks: Media Independent Handover Services. May 2008. [44] S. Goudarzi, W.H. Hassan, M.H. Anisi, S.A. Soleymani, P. Shabanzadeh, A novel model on curve fitting and particle swarm optimization for vertical handover in heterogeneous wireless networks, Mathematical Problems in Engineering. 2015 (2015) 1–16. [45] S. Goudarzi, W.H. Hassan, S.A. Soleymani, O. Zakaria, L.B. Jivanadham, Artificial bee colony for vertical-handover in heterogeneous wireless networks, Advanced Computer and Communication Engineering Technology, Springer International Publishing, 2016, pp. 307–322. [46] D. Karaboga, B. Basturk, A powerful and efficient algorithm for numerical function optimization: artificial bee colony (abc) algorithm, J. Global Optim. 39 (2007) 459–471. [47] D. Karaboga, B. Basturk, On the performance of artificial bee colony (abc) algorithm, Appl. Soft Comput. 8 (2008) 687–697. [48] R. Eberhart, J. Kennedy, A new optimizer using particle swarm theory, in: Proceeding of Sixth International Symposium on Micro Machine and Human Science, 1995, pp. 39–43. [49] A. Ahmed, L. Merghem-Boulahia, D. Gaïti, An intelligent agent-based scheme for vertical handover management across heterogeneous networks, Ann. Telecommun. 66 (9-10) (2011) 583–602. [50] S. Goudarzi, W.H. Hassan, M.H. Anisi, S.A. Soleymani, Comparison between hybridized algorithm of GA–SA and ABC, GA, DE and PSO for vertical-handover ¯ in heterogeneous wireless networks, Sadhan a¯ 41 (1998) 727–753. [51] T.L. Saaty, What is the analytic hierarchy process?, Mathematical Models for Decision Support, Springer, Berlin Heidelberg, 1988, pp. 109–121. [52] C. Hue, J.P. Le Cadre, P. Perez, Sequential Monte Carlo methods for multiple target tracking and data fusion, IEEE Trans. Signal Process 50 (2) (2002) 309–325. [53] A. Doucet, N. De Freitas, N. Gordon, An introduction to sequential Monte Carlo methods,, Sequential Monte Carlo Methods in Practice, Springer, New York, 2001, pp. 3–14. [54] L. Mihaylova, D. Angelova, S. Honary, D.R. Bull, C.N. Canagarajah, B. Ristic, Mobility tracking in cellular networks using particle filtering, IEEE Trans. Wireless Commun. 6 (10) (2007) 3589–3599. [55] J.H. Kotecha, P.M. Djuric´ , Gaussian sum particle filtering, IEEE Trans. Signal Process. 51 (10) (2003) 2602–2612. [56] J.M. Marquez-Barja, H. Ahmadi, S.M. Tornell, C.T. Calafate, J.C. Cano, P. Manzoni, L.A. DaSilva, Breaking the vehicular wireless communications barriers: vertical handover techniques for heterogeneous networks, IEEE Trans. Veh. Technol. 64 (12) (2015) 5878–5890.

[57] C. Guo, Z. Guo, Q. Zhang, W. Zhu, A seamless and proactive endto-end mobility solution for roaming across heterogeneous wireless networks, IEEE J. Selected Areas Commun. 22 (5) (June 2004) 834–848. [58] O. Ormond, J. Murphy, G. Muntean, Utility-based Intelligent Network Selection in Beyond 3 G Systems, in: Proceedings of IEEE ICC’06, Istanbul, Turkey, June 20 06, 20 06, pp. 947–951. [59] E. Stevens-Navarro, V.W. Wong, Comparison between vertical handoff decision algorithms for heterogeneous wireless networks, in: 2006 IEEE 63rd Vehicular Technology Conference, Vol. 2, IEEE, 2006, May, pp. 947–951. [60] E. Navarro, W. Wong, L. Xia, A vertical handoff decision algorithm for heterogeneous wireless networks, in: IEEE Wireless Communication Network Conference (WCNC), 2007, pp. 1–6. [61] F. Bari, V. Leung, Automated network selection in a heterogeneous wireless network environment, IEEE Netw 21 (1) (2007) 34–40. Shidrokh Goudarzi is a Ph.D. student in Universiti Teknologi Malaysia. Her field of study is Communication System and Wireless Network. She received the B.Sc. in Computer Engineering (Software Engineering) degree from Sadjad University of Iran. She has academic experience from Islamic Azad University in Iran. Her research interests are in wireless networks, artificial intelligence, and next generation networks.

Wan Haslina Hassan is an Associate Professor of Computer Communications Computer Networks. She obtained her Ph.D. from Universiti Teknologi Malaysia (UTM). Her research interests lie in the area of Mobile Computing, Intelligent Architectures and Network Security.

Mohammad Hossein Anisi is a senior research associate at the School of Computing Sciences, University of East Anglia. He obtained his PhD from Universiti Teknologi Malaysia (UTM) while being awarded as the best postgraduate student. He worked as post-doctoral research fellow at UTM and was a member of pervasive computing research group (PCRG), a research group under KEconomy Research Alliance in Malaysia. He also worked as senior lecturer at the Department of Computer System and Technology, Faculty of Computer Science and Information Technology, University of Malaya. His research interests lie in the area of internet of things, wireless sensor networks and their applications, mobile ad hoc networks and intelligent transportation systems. He has published several papers in international journals and conferences and secured a number of national and international research grants. He has also collaborated actively with researchers in several other disciplines of computer science. Seyed Ahmad Soleymani is a Ph.D. candidate of the Department of computing, Universiti Teknologi Malaysia (UTM), Malaysia. He received the M.S. from the Department of Computer Engineering, Islamic Azad University, Iran and B.S. degree from the Department of Computer Engineering, Sadjad University, Iran. His research interests include Routing, Trust, and Security in WSN, VANET security and Trust in VANET.

Mehdi Sookhak is a postdoctoral fellowship at Carleton University of Canada funded by Canadian Natural Sciences and Engineering Research Council (NSERC). He was an active researcher in Center of Mobile Cloud Computing Research (C4MCCR) at Faculty of Computer Science and Information Technology, University Malaya, Kuala Lumpur. His areas of interest include Cryptography and Information Security, Mobile Cloud Computing, Computation outsourcing, Access control, Wireless Sensor & Mobile Ad Hoc Network (Architectures, Protocols, Security, and Algorithms), and Distributed Systems.


JID: NEUCOM


Muhammad Khurran Khan is currently working as a Full Professor at the Center of Excellence in Information Assurance (CoEIA), King Saud University, Kingdom of Saudi Arabia. He is one of the founding members of CoEIA and has served as the Manager R&D from March 2009 to March 2012. He developed and successfully managed the research program of CoEIA, which transformed the center as one of the best centers of research excellence in Saudi Arabia as well as in the region. Prof. Khurram has published over 250 research papers in the journals and conferences of international repute. In addition, he is an inventor of 10 US/PCT patents. He has edited 7 books/proceedings published by Springer-Verlag and IEEE. He has secured several national and international research grants in the domain of information security. His research areas of interest are Cybersecurity, digital authentication, biometrics, multimedia security, and technological innovation management. Prof. Khurram has recently played a leading role in developing ’BS Cybersecurity Degree Program’ and ’Higher Diploma in Cybersecurity’ at King Saud University.

[m5G;March 16, 2017;4:31] 19

Aisha-Hassan Abdalla Hashim is currently an associate professor at the Department of Electrical and Computer Engineering, Faculty of Engineering, International Islamic University Malaysia. She received her BSc and MSc in Computer Engineering from University of Gezira, Sudan, and her Ph.D. in Computer Engineering from International Islamic University Malaysia (IIUM) in 2007. She was appointed as a lecturer in 1997. Her current research interests include Data Communication and Computer Networking, ASIC Design, Computer Architecture and Grid Computing, Open Sources & Operating Systems. She published more than 100 papers in international journals and conferences. She is IEEE Senior Member, IEEE Women in Engineering Member and works as a reviewer for many ISI journals. Mahdi Zareei received his B.E. degree in Computer Engineering from Payame Noor University of Hamedan, Iran in 2008 and respectively M.Sc., degree in Computer Network from Universiti Sains Malaysia in 2012. Currently he is a Ph.D. student in Malaysia-Japan International Institute of Technology (MJIIT), Universiti Teknologi Malaysia (UTM). His research interest includes wireless sensor network, energy efficient protocols, distributed computing.
