ABSTRACT

In recent years, telecommunication has flourished extensively so much that several areas of studies coexist now thanks to multiple technologies. One of them is heterogeneous handover which is a concept that aims to provide continuity of connection while crossing different networks.

In this thesis, our main objective is to analyze the handover between two WLAN, two Wimax and two UMTS networks. The vertical handover decision is taken on the basis of various algorithms such as variance-based algorithm, taguchi algorithm, which calculates the variance of parameters such as delay, jitter, bandwidth and packet loss for the above networks, and selection of the network having most parameters with minimum score. These algorithms are calculated and the decision factors for each wireless network are compared, in order to detect and trigger a vertical handover. The factors can be classified as beneficial, i.e., the larger, the better, or on the basis of cost, i.e., the lower, the better.

This algorithm is also compared with other algorithms such as MEW (Multiplicative experiment weighting), SAW (Simple Additive Weighting), TOPOSIS (Technique for order preference by similarity to ideal solution) and GRA (Grey Relational Analysis). These algorithms are appropriate for different traffic classes. Simulation results for the proposed algorithm in Matlab is discussed and compared with other multiple attribute decision making algorithms on the basis of bandwidth, jitter, delay etc. It can be seen that the proposed algorithm causes the minimum packet delay than others. Jitter is also comparatively less than other algorithms. Besides, it provides the highest bandwidth than any other MADM algorithm.

Keywords:-UMTS, SAW, MEW, GRA, TOPSIS, WLAN and Wimax

Chapter-1

Introduction

1.1 History of mobile services ‘ The journey of mobile telephony began with the 1st generation services. The design for it was developed by AMPS (Advanced mobile phone system) in 1970 and it is based on analog cellular technology. The data bandwidth provided by the system was just 1.9 kbps and it used TDMA multiplexing. Then, the 2nd generation of mobile services was introduced in 1981.The 2G systems are still largely used for voice calls. The data bandwidth provided was 14.4 kbps [1] and the technology used was TDMA and CDMA. It is based on digital technology and also provided short messaging services or SMS along with voice communication. Similarly, it provided circuit switched data communication services at low speed. In 1999, the technology switched to 2.5 G, which used GPRS, EDGE as the standards. It provided higher throughput for data service up to 384 kbps. Later, in 2002, the 3G services were introduced, providing high quality audio, video and data services. Which also provided broadband data capabilities up to 2 Mbps. It mainly uses packet switched technology which utilizes the bandwidth more efficiently. In 2010, when 4th generation of cellular technology was introduced, it was expected to complement and replace the 3G networks. The key features of 4G mobile networks is that it can deliver information anywhere and anytime using seamless connection.4G network is an IP based network which gives access through collection of various radio interfaces. Its network provides access to best possible service with seamless handoff, combining multiple radio interfaces into a single network for subscriber to use. Thus, users have different services with an increased coverage. It does not matter whether there is failure or loss of one or more networks, the 4G technology keeps all the networks integrated into IP based system, which require vertical handoff for seamless connection between the networks. As the number of users are responsible for enhancing the quality of 4G service, the very process becomes an indispensable component. While the 4G technology has its genesis in the idea of invasive computing, software defined radio is the prominent adhesive behind the entire process. Here the software defined radio is programmable and able to transmit and receive a wide range of frequencies while emulating any transmission format. It should offer high speed of 100 Mbits for stationary mobile and 20 Mbits while travelling having network capacity 10 times faster than 3G networks.

This increases the download speed to 1 second for 1Mbyte of file compared to 200 seconds in 3G networks. Which should support fast speed volume data transmission at lower cost. The obviously it should provide seamless connection between multiple wireless networks and mobile networks. For this, the support of vertical handoff is essential. Apart from it, it is expected that seamless multimedia services are provided it being an IP-based system, which also replaces SS7 (signaling system 7) that consumes considerable amount of bandwidth. Due to IP-based network, optimum usage of bandwidth is expected.

1.2 Motivation ‘

There are several communication systems such as the Ethernet, Wireless LAN, GPRS and 3G coexisting with their own different characteristics such as bandwidth, delay and cost.

Wireless mobile users require high quality of service (QoS) and one of the factors directly affecting QoS is the number of call drops. Therefore, it has to be reduced or eliminated, possibly, to achieve high QoS. The number of call drops experienced by a system mainly depends on its channel assignment and handoff schemes. Since majority of WLANs are deployed in the areas like hotels, cafes, airports, offices, etc, the speed of the users are generally normalized within the WLAN coverage area. In WLAN/Cellular network interworking, a user can either have access from micro layer or from macro layer of cellular network depending whether he wants slow or fast speed. Basically, in cellular networks, user speed is the primary factor to determine whether a user is fast or slow and that information is subsequently used to handle vertical handoff. Since the speed information about the users are not directly available when they are in WLAN coverage area ,the vertical handoff schemes employed in cellular networks are not directly suitable for solving vertical handoff problem in Cellular/WLAN interworking. This raises to many important questions.

Assume that the speed of each user in a WLAN coverage area is within small threshold value. Normally, when a user is outside WLAN coverage area, the type of user whether fast or slow, is determined on the basis of the user speed. Now the question is how to determine the type of user, whether fast or slow, when it is within the WLAN coverage area?

A fast user can become slow temporarily due to various conditions such as traffic signals, turns, etc. Is the speed alone sufficient to determine whether a user is slow or fast? If not then (i) What other parameters can be used to determine whether a user is fast or slow and,

ii) How can they be obtained?

In fact, this lack of clarity inspired the researcher to develop a vertical Handoff decision algorithm to solve vertical handoff problems.

1.3 Scope of thesis ‘

With greater mobility and easy acces, telecommunication consumers have become demanding, seeking services anywhere and anytime. Thus, the integration of WLAN (Wireless LAN), Wimax and cellular networks such as WCDMA (wideband CDMA) system should be error free for seamless efficient communication which is the 4th generation technology. The seamless and efficient handover between different access technologies known as vertical handover is essential and remains a challenging problem. The 4G is seen as convergence and integration of various wireless access technologies. The existing cellular systems such as GSM and CDMA2000 support low bandwidth over a large coverage area. However, the wireless networks such as WLAN supports high bandwidth over a short coverage area. Moreover one of the major design issues of 4G is the support of vertical handover.

Interestingly this is different from a ‘horizontal handoff’ between different wireless access points that use the same technology. Switching between two dissimilar networks for mobile terminal (e.g. between UMTS & WLAN) is termed as Vertical Handover

A vertical handover involves two different network interfaces for different wireless technologies. It can happen in two ways. Firmly when the mobile user moves into the network that has higher bandwidth and limited coverage, a vertical handover request is generated since the mobile user may want to change its connection to the higher bandwidth network to enjoy the higher bandwidth service. This type of vertical handover is called downward vertical handoff. Secondly when the mobile user moves out of its serving higher bandwidth network, it has to request a vertical handover to change its connection to the network with low bandwidth and wide coverage. This type of vertical handover is called upward vertical handover.

Chapter-2

Research objectives

2.1 Objectives

The present research aims at making comparison between various existing multiple attribute decision making algorithms for realization of vertical handoff such as MEW (Multiplicative Exponent Weighting), SAW (Simple Additive weighting), TOPSIS (Technique for order preference by similarity to ideal solutions) and GRA (Grey relational Analysis) which are MADM (Multiple attribute decision making) ranking algorithms and the proposed vertical decision algorithm.

For this comparison, various heterogeneous networks such as UMTS (Universal Mobile Telecommunication services), WLAN (Wireless Local area networks), WiMAX (Worldwide interoperability for microwave access) need to be taken into consideration.

Comparison will be mostly on the basis of various parameters such as bandwidth, jitter, packet delay, packet loss, etc. In addition, the comparison may be for different types of traffics such as data connections and voice connections. As all the above mentioned algorithms are multiple attribute algorithm, due importance is given to parameters to be considered in the algorithms. The performance evaluation of the proposed decision algorithm should be done on the basis of parameters mentioned above. For various types of traffics, how the algorithm performs can be seen. Depending on the performance, we can conclude Which algorithm is suitable for which traffics. In voice connections, 70 % importance is given to the packet delay and jitter i.e. by assigning weights to these parameters and equal distribution of weights to the other parameters or attributes. If any of the algorithms performs well then that particular algorithm can be considered to be best suited for voice connections. In data connections, 70% importance is given to the parameters such as bandwidth i.e. by assigning the weight to the bandwidth and remaining weight is equally distributed among the parameters. If any of the algorithm performs well in this case, then the particular algorithm is suitable for the data connections. The ultimate aim being development of a decision making algorithm which works well for both voice connections and data connections.

‘

2.2 Methodology

In order to realize vertical handoff using the existing multiple attribute decision making algorithm and evaluate the performance of each of the algorithms along with the proposed algorithm, we are considering the selection of network in 4G environment. Here, three types of networks such as UMTS (Universal Mobile Telecommunication services), WLAN (Wireless Local area networks), WiMAX (Worldwide interoperability for microwave access) are combined and there will be two networks of each type. In this thesis, four decision criteria are evaluated and compared to realize vertical handoff considering the available bandwidth (Mbps), packet delay (ms), packet jitter and packet loss (per 106 packets). The range of value for various parameters are as follows: Available bandwidth for UMTS network 0.1-2Mbps, Packet delay for UMTS network 25-50ms, Jitter for UMTS network 5-10ms. Bandwidth for WLAN network 1-54Mbps, Packet delay for WLAN network 100-150ms, Jitter for WLAN network 10-20ms.

Bandwidth for Wimax network is 1-60Mbps, while for packet delay for Wimax network is 60-100ms, and Jitter for Wimax network is 3-10ms.The values for the weights to be assigned for different services are considered as Case1: packet delay and jitter are given 70 % importance and the rest is equally distributed among other parameters, this case is suited for voice connections and whereas Case 2: available bandwidth is given as 70% importance, this case is suited for data connections. For each algorithm, 10 vertical decisions were considered of each case separately. Performance evaluation is done for two cases namely voice connections and data connections. These cases are evaluated using MATLAB v7.6 release 2009 software tool.

Next, by using artificial neural network, we can design a system to take vertical handoff decision. Here, input parameters such as samples of received signal strength and bandwidth is applied to input layer, hidden layer does some processing depending upon the number of neurons and the algorithm chosen. The output layer gives the ID of selected candidate network. In, ANN-based method, there is handoff between WLAN and Cellular networks. Here, two parameters are taken into consideration i.e. RSS a Bandwidth as an input for neural network. The RSS samples for training neural network for both WLAN & cellular networks are -60dBm,-70 dBm,-80 dBm,-90 dBm. Similarly, bandwidth samples for WLAN are 54, 30,10,1 Mbps. Bandwidth samples for cellular network are 14.4, 9.6, 4.5,2 kbps. By using combination of RSS & bandwidth parameters, we could make 256 samples of input for ANN. These samples of output samples for vertical handoff decision are also fed to ANN.

Using Levenberg-Marquardt method for ANN, 180 samples are used for training, 38 samples for validation and 38 samples for testing. Based on ANN developed system, it could take vertical handoff decision from cellular to WLAN. Lastly, ns-2 software tool can also be used. NS-2 simulation is done using nodes of 802.11 and nodes of 802.16 Wimax. Four nodes of 802.11 nodes (Access points) are used and four nodes of 802.16 nodes (Base station) NIST module of 802.16e are used in ns-2. In this case, out of the existing algorithm best algorithm with best score is selected for triggering vertical handoff decision. Here, in this case, various parameters such as Bandwidth, Bit error rate, trust level etc were considered for vertical handoff decision. This can be tested against various available traffics in ns2 such as CBR (Constant bit rate) which corresponds to real time traffic (for voice communication) and FTP (file transfer Protocol) which corresponds to non real time traffic. The performance evaluation for various traffics can be done on various parameters such as Packet delivery ratio, throughput, jitter and packet dropping ratio etc with simulation time.

2.3 Related Work

Enrique Stevens Navarro and Vincent W.S.Wong [2], in their paper, have compared four different vertical handoff decision algorithm namely, MEW (Multiplicative Exponent Weighting), SAW (Simple Additive Weighting), TOPSIS (Technique for Order Preference by Similarity to Ideal Solution), and GRA (Grey Relational Analysis). All four algorithms allow different parameters (e.g., bandwidth, delay, packet loss rate, cost) to be considered for vertical handoff decision [2]. Both Authors found that MEW, SAW, and TOPSIS provide almost the same performance to all four traffic classes. Only GRA gives a slightly higher bandwidth and lower delay for interactive and background traffic classes.

Jose.D.Martinez, Ulises Pinedo-Rico and Enrique Stevens Navarro, in their paper, have given a comparative analysis of the multiple attribute decision algorithms [3]. In this paper, the authors provided a simulation study of several vertical handoff decision algorithms in order to understand its performance for different user applications. They considered two different applications: voice and data connections. Algorithms such as SAW (Simple Additive Weighting) and TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) are suitable for voice connections. These algorithms provided the lower values of jitter and delay packet available in a 4G wireless network. In a data connection case, GRA (Grey Relational Analysis) and MEW (Multiplicative Exponent Weighting) algorithms provided the solution with highest available bandwidth necessary for this application.

Chapter-3

Classification of vertical handoff algorithms

3.1 Need for vertical handoff

Currently, the trend in mobile communications is not one network technology replacing another, but the interoperability between different overlapping networks. Therefore it is obvious that many wireless networks will coexist and can complement each other in an all-IP based heterogeneous wireless network. This can facilitate mobile users’ access to internet easily and connectivity of IP anywhere, anytime using the ‘best’ possible network. This is mainly due to the fact that different wireless networking technologies have their own advantages and drawbacks. Access to various wireless systems results in heterogeneous networks that can offer overlapping coverage of multiple networks with different technologies. For example, low cost and high speed Wi-Fi (WLAN) network will be accessible within limited range of ‘hot-spot’ areas and will be complimented with cellular network offering wide area coverage such as UMTS or Wimax. As a consequence, some fundamental problems must be solved for the users to navigate a 4G wireless network seamlessly. For this, mobile terminal equipped with multiple interfaces to handle different technologies is required. Furthermore, applications running on mobile terminal with multi-mode terminals in a 4G environment can switch between different networks supporting different technologies without degrading the quality of the link. But the Internet routing model forces mobile terminal to find new IP address for an interface while roaming in another network in 4G environment. It is assumed that applications can easily manage mobility and can handoff to the best possible network; of course some method is required to adjust media streams to the bandwidth available.

3.1 Types of vertical handoff

There are various ways to classify vertical handoff algorithms. In this thesis, we have classified the vertical handoff algorithms into four groups based on the handoff criteria as given below:

RSS-based algorithms: RSS is used as the main handoff decision criteria in this group. Different strategies have been developed to compare the RSS of the present point of attachment. In this RSS-based horizontal handoff decision, strategies are classified into the following six subcategories namely: relative received signal strength, relative received signal strength with threshold, relative received signal strength with hysteresis and threshold, and prediction techniques. For vertical handoff decision, relative received signal strength cannot be applicable since the signal strength from different types of networks cannot be compared directly due to the different technologies involved. For example, different thresholds for different networks. Furthermore, other network parameters such as bandwidth are combined with RSS in the vertical handoff decision process.

Bandwidth-based algorithms: Available bandwidth for a mobile terminal is the main criterion in this group. In some algorithms, both bandwidth and signal strength information are used in the decision process. Depending on whether RSS or bandwidth is the main criteria an algorithm is categorized either as signal strength based or bandwidth based.

Cost function based algorithms: This class of algorithms combine metrics such as monetary cost, security, bandwidth and power consumption in a cost function based algorithm, and the handoff decision is made by comparing the score of this function for the candidate networks. Different weights are allotted to the different input parameters depending on the network conditions and user preferences.

ANN and fuzzy logic based algorithms: These vertical handoff decision algorithms attempt to use richer set of inputs than others for making handoff decisions. When a large number of inputs are used, it is very difficult or impossible to develop, formulate handoff decision processes. Analytically hence, it is better to apply machine learning techniques to formulate the processes. The survey reveals that for the fuzzy logic and artificial neural networks based techniques can be used. The Fuzzy logic systems allow expertise of humans for qualitative thinking to be incorporated as algorithms to enhance the efficiency. If there exists comprehensive set of input-desired output pattern, artificial neural networks can be trained to create handoff decision algorithms. By using consistent and real-time learning techniques, the systems can monitor their performance and change their own structure to create very effective handoff decision algorithms.

3.3.1 RSS based vertical handoff

In this, the handoff decisions are made by comparing RSS (received signal strength) of the current network with the preset threshold values. These algorithms are less complex and may be combined with other parameters such as bandwidth, cost etc to have a better handover decisions. We describe here three RSS based algorithms in the following sections.

A) ALIVE-HO (adaptive lifetime based vertical handoff ) algorithm – Zahran, Chen and Sreenan proposed algorithm for handover between 3G networks and WLAN by combining the RSS with an estimated lifetime (duration over which the current access technology remains beneficial to the active applications). ALIVE-HO always uses an uncongested network whenever available. It continues using the preferred network (i.e. WLAN) as long as it satisfies the QoS requirements of the application [5].

Two different vertical handoff scenarios let us discuss: Moving out of the preferred network (MO) and Moving in to the preferred network (MI), where the preferred network is usually the underlay network that provides better and economical service. Hence, extending the utilization of WLAN as long as it provides satisfactory performance is the main consideration of vertical handoff algorithm design. We observe the method through the following scenarios. In the first scenario, when the MT moves away from the coverage area of a WLAN into a 3G cell, a handover to the 3G network is initiated. The handover is done under following conditions:

(a) RSS average of the WLAN falls below predefined threshold. (MO threshold) and (b) the estimated lifetime is at least equal to the required handoff signaling delay. The MT continuously calculates the RSS mean using the moving average method.[4]

[K] =

Here [k] is RSS mean at time instant k, and Wav is the window size, a variable that changes with velocity of the velocity of mobile terminal. Then, the lifetime metric EL [k] is calculated by using [k], ASST Application signal strength threshold),S[k],RSS change rate.

EL[k] = [k] ‘ ASST

S[k]

ASST (Application signal strength threshold) chosen to satisfy the requirements of the active applications. S[K] represents RSS decay rate. In second scenario, when the MT moves towards a WLAN cell, the handover to the WLAN is done if the average RSS is larger than MI Threshold. WLAN and the available bandwidth of the WLAN meet the bandwidth requirement of the application. Table 3.1 given below shows lost frames during the handoff transition area for the received stream.

ASST (in dBs) -90 -89 -88 -87 -86 -85

Lost frames_100kbit/s 13.3 5 3 0.67 0 0

Lost frames_300kbit/s 38 28 4 0.33 0 0

Table 3.1 Frames lost corresponding to ASST [5]

Based on the results obtained and subjective testing, the optimal value for UDP based streaming is chosen as -86dB.

Firstly by introducing EL[k], the algorithm adapts to the application requirements and reduces unnecessary handovers. Secondly, there is an improvement on the average throughput for user because MT prefers to stay in WLAN cell as long as possible.

However, packet delay grows due to the critical fading impact near the cell edges, which may result in severe degradation in the user perceived QoS. This phenomenon results in a tradeoff between improving the system resource utilization and satisfying the user QoS requirements. This issue can be critical for delay sensitive applications and degrade their performance. Here ASST is tuned according to various system parameters, including delay thresholds, MT velocities, handover signaling costs and packet delay penalties.

B) Algorithm on Adaptive RSS Threshold

Mohanty and Akyildiz, in their paper, have proposed a WLAN to 3G handover decision method [6]. In this method, RSS of current network is compared with dynamic RSS threshold (Sth) when MT is connected to a WLAN access point. We observe the following notations with reference to fig 3.1 which shows a handoff from current network (AP) referred as WLAN, to the future network (BS), referred as 3G.

Fig 3.1 Analysis of handoff process [6]

* Sth: The threshold value of RSS to initiate the handover process. Therefore, when the RSS of WLAN referred to as ORSS (old RSS) in fig 3.1 drops below Sth, the registration procedures are initiated for MT’s handover to the 3G network.

* a:The cell size we assume that the cells are of hexagonal shape.

d: It is the shortest distance between the point at which handover is initiated and WLAN boundary. We observe the Path loss Model given by;

Pr(x) = Pr (d0) +

Where x is the distance between the Access Point and Mobile Terminal, and Pr (d0) is the received power at a known reference distance (d0). The typical value of d0 is 1 km for macrocells, 100m for outdoor microcells, and 1m for indoor picocells. The numerical value of Pr (d0) depends on different factors such as frequency, antenna heights, and antenna gains, is the path loss exponent. The typical values of ranges from 3 to 4 and 2 to 8 for macrocellular and microcellular environment.

– Is a Zero mean Gaussian random variable that represents the statistical variation in Pr(x) caused by Shadowing. Typical std. deviation of is 12 dB.

We observe the path loss model applied to the scenario.

Pr (a ‘ d) = Pr (a) +

Pr (a ‘ d) = Pr (a) + 10 log

Sth = Smin + 10 log

When the MT is located at point P, the assumption is that it can move in any direction with equal probability, i.e. the pdf of MT’s direction of motion is

f = – < < ”.’.(1)

As per assumption, that MT’s direction of motion and speed remains the same from point P until it moves out of the coverage area of WLAN. As the distance of P from WLAN boundary is not very large, this assumption is realistic.

The need for handoff to cellular network arises only if MTs direction of motion from P is in the range [ ] can move in both directions.

Where = arctan otherwise the handoff initiation is false. The probability of false handoff initiation is using (1) is

Pa = 1 –

P (unfavourable event ) = 1 ‘ P ( favourable event )

= 1 –

= 1 – ”.’.(2)

When the direction of motion of Mobile Terminal from P is, the time it takes to move out of the coverage area of WLAN cell i.e. old base station is given by

time =

from fig 3.1 Cos =

Sec = , x = d sec

Hence t =

t= ”.’.(3)

Pdf of is

f = “{” 1/(2”_1 ) -”_1<”<”_1 {0 otherwise from (3) , t is a function of i. e. t = g ( ) in [ ] [3] g(”)=dsec”/v Therefore pdf of t is given by f_t (t)=’_i’f_(” (”_i ) )/(g^’ (”_i ) ) ”.’.(4) Where ”i are the roots of equation t = g ( ) in [ ] And for each of these roots f”(”i)= for i = 1 and 2 f = + f = ”.’.(5) Where g is derivative of g given by g = ”.’.(6) = = g = t Using (5) & (6), the pdf of t is given by f = ”.’.(7) { 0 otherwise The probability of handoff failure is given by Pf = { 1 >

{ P ( t < ) < <

{ 0 ”.’.(8)

– handoff signaling delay

and P ( t < ) – is the probability that t <

when

P (t < ) = = = = arccos( ) ”.’.(9) Using (8) and (9) we get Pf = { 1 >

{ cos < <

{ 0

Pf =

Since, ”1 =arctan ( d/v”)

Pf =

Here, it shows that Probability of handoff failure depends on distance from point p to the boundary of the cell, velocity and handoff signaling delay ”.

The use of adaptive RSS threshold helps reducing the handoff failure probability as well as reducing unnecessary handovers. The exact value of Sth will depend on MT’s speed and handoff signaling delay at a particular time. Adaptive Sth is used to limit handoff failure. However, in this algorithm, the handoff from 3G network to a WLAN is not efficient when MTS traveling time inside a WLAN cell is less than the handover delay. This may lead to wastage of network resources.

3.3.2 Bandwidth based vertical handoff algorithm

A Signal to Interference and Noise Ratio (SINR) Based algorithm

Yang, in his paper,[7] presented a bandwidth based vertical handover decision method between WLANs and a Wideband Code Division Multiple Access (WCDMA) network using Signal to Interference and Noise Ratio (SINR) algorithm[7]. The SINR calculation of the WLAN (wireless LAN) signals is converted into an equivalent Signal to Interference and noise Ratio to be compared with the Signal to Interference and noise Ratio of the Wideband Code Division Multiple Access channel

”AP =”AP [(1+ ”BS/ ”BS) WBS/ WAP -1]

where ”AP and ”BS are the Signal to Interference and noise Ratio at the mobile terminal when associated with Wireless local area network and Wideband Code Division Multiple Access, respectively. ” is the dB gap between the uncoded Quadrature Amplitude Modulation and channel capacity, minus the coding gain, and ”AP equals to 3dB for Wireless local area network and ”BS equals to 3dB for Wireless local area network, as stated by the authors. WAP and WBS are the carrier bandwidth of wireless local area network and Wideband Code Division Multiple Access links. Signal to Interference and Noise Ratio based handovers can provide users with higher overall throughput than RSS based handovers since the available throughput is directly dependent on the Signal to Interference and Noise Ratio, and this algorithm results in a balanced load between the wireless local area network and the Wideband Code Division Multiple Access networks. But such an algorithm may also introduce excessive handovers with the variation of the Signal to Interference and Noise Ratio causing the node to hand over back and forth between two networks, commonly referred to as ping-pong effect.

A Wrong Decision Probability (WDP) Prediction Based algorithm

C.Chi, Cao, Hao and Liu, in their paper ‘Modeling and analysis of Handover algorithms’, have proposed a Vertical Handover decision algorithm based [8] on the Wrong Decision Probability (WDP) prediction. The Wrong Decision Probability is calculated by combining the probability of unnecessary handoff and the missing handoff. Assume that there are two networks i and j with overlapping coverage, and bi and bj are their available bandwidth. An unnecessary handoff occurs when the mobile terminal is in network i and decides to handoff to j, but bj is less than bi after this decision. A missing handoff occurs when the mobile terminal decides to stay connected to network i, but bi is less than bj after this decision. A handover from network i to network j is initiated if Pr < ” x l0 or bj – bi ‘ L, where Pr is the unnecessary handover probability, ” is the traffic load of network i, l0 = 0.001, and L is a bandwidth threshold. The authors show that this algorithm is able to reduce the Wrong Decision Probability and balance the traffic load; however, received signal strength is not considered. A handoff to a target network with high bandwidth but weak received signal is not desirable as it may bring discontinuity in the service.

3.3.3 Cost Function based vertical handoff algorithm

A Cost Function Based algorithm with Normalization and Weights Distribution

Hasswa, N. Nasser, and H. Hassanein, in their paper ‘A context-aware cross-layer archi-

tecture for next generation heterogeneous wireless networks’, have proposed a cost function based handover decision algorithm in which the normalization and weights distribution methods are provided. A quality factor of network is used to evaluate the performance of a handover target candidate as

Qi = WCCi + WSSi + WPPi + WdDi + WfFi

where Qi is the quality factor of network i, Ci, Si, Pi, Di and Fi stand for cost of service, security, power consumption, network condition and network performance, and Wc, Ws, Wp, Wd andWf are the weights of these network parameters. Since each network parameter has a different unit, a normalization procedure is used and the normalized quality factor for network n is calculated as

Wc(1/Ci) WSSi WP(1/Pi)

Qi = ”__________ + __________ + _______

max((1/C1),’..(1/Cn)) max(S1,’..Sn) max((1/P1),’..(1/Pn))

WdDi WfFi

+ ________ + ________

max (D1,’..Dn) max(F1,’..Fn)

A handoff necessity estimator is also introduced to avoid unnecessary handovers High system throughput and user’s satisfaction can be achieved by introducing Hasswa’s algorithm, however, some of the parameters such as security and interference levels are difficult to calculate.

A Weighted Function Based Algorithm

R. Tawil, G. Pujolle, and O. Salazar in their paper presented a weighted function based[10] Vertical handover decision algorithm which transfers the Vertical handover decision calculation to the visited network instead of the mobile terminal. The weighted function of a candidate network is defined as

Qi = WBBi +WDp 1/DPi+WC 1/Ci

Where Qi represents the quality of network i, Bi, DPi and Ci are bandwidth, dropping

probability and monetary cost of service, and WB, WDp and WC are their

weights, where,

WB +WDp +WC = 1

The candidate network with the highest score of Qi is selected as the handover target. By giving the calculation to the visited network, the resource of the mobile node can be saved so that the system is able to achieve short handoff decision delay, low handoff blocking rate and higher throughput. However, the method requires extra cooperation between the mobile node and the point of attachment of the visited network, which may cause additional delay and excessive load to the network when there are large number of mobile nodes.

3.3.4 ANN based vertical handoff algorithm

A Multilayer Feedforward Artificial Neural Network Based Algorithm

N. Nasser, S. Guizani, and E. Al-Masri, in their paper, developed a [11] Vertical handover decision algorithm based on artificial neural networks (ANN). The topology of the ANN consists of an input layer, a hidden layer and an output layer. The input layer consists of five nodes representing various parameters such as cost, RSS, bandwidth etc of the handoff target candidate networks. The hidden layer consists of variable number of nodes (neurons) which are basically activation functions. The output layer has one node which generates the ID of the candidate network of the handover target. All the neurons use sigmoid activation function. The authors have assumed the same cost function as in this work and also for ANN training they have generated a series of user preference sets with randomly selected weights. Then the system has to be trained to select the best candidate network among all the candidates. The authors have reported that by properly selecting the learning rate and the acceptable error value, the system is able to find the best available candidate network successfully. Nevertheless, the algorithm suffers from a long delay during the training process which may lead to connection breakdown.

A Method That Uses Two Neural Networks

Pahlavan, in his paper, has proposed two neural [12] network based decision methods of vertical handoff. Here, only the vertical handoff mechanism is discussed. In the method for vertical handoff, an artificial neural network is used for handoffs from the Wireless local area network to the General Packet Radio Service (GPRS). The Artificial neural network consists of an input layer, two middle layers and an output layer. Mobile node does periodical measurements of RSS and five most recent samples of RSS are fed into the ANN. The output is a binary signal: The value ‘1’ leads to a handover to the General Packet Radio Service, and the value ‘0’ means that the mobile terminal should remain connected to the access point. The ANN is trained before it is used in the decision process. Training is done by taking a number of RSS samples from the access point while minimizing the handover delay and ping-pong effect. This algorithm can reduce the number of handovers by eliminating the ping-pong effect, but the paper does not provide details on how exactly the neural network is trained and why the particular parameters are taken into consideration. This algorithm also has the short coming of the algorithm complexity and the training process to be performed in advance.

Summary:

From the above discussion, it can be concluded that RSS based Vertical handoff algorithms can be used between microcellular and macro cellular networks. The network candidate with most stable RSS being the selection criteria. These algorithms are simple, but due to the fluctuation of RSS, they are less reliable.

Bandwidth based Vertical handoff algorithms can be used between any two heterogeneous networks. The network candidate with highest bandwidth is the selection criteria. These algorithms are simple. But, due to the changing available bandwidth, these algorithms are less reliable.

Cost function based Vertical handoff algorithms can be used between any two heterogeneous networks. Here, the inputs are various parameters such as cost, bandwidth, security etc The network candidate with highest overall performance is the selection criteria. These algorithms are complex. But, due to the difficulty in measuring parameters such as security etc, they are less reliable.

ANN and Fuzzy logic based Vertical handoff algorithms can be used between any two heterogeneous networks. Here, the inputs are various parameters such as RSS, cost, bandwidth, security etc depending on different methods. The network candidate with highest overall performance is the selection criteria. These algorithms are very complex. But, due to training of system, they are highly reliable.

‘

Chapter-4

Algorithms and Methodologies

4.1 Variance based vertical handoff algorithm

Proposed algorithm is variance based algorithm which calculates the variance of parameters such as delay, jitter, bandwidth and packet loss for various networks such as UMTS,WLAN,Wimax and the network with most of the parameters with minimum variance being selected. In our proposed algorithm, handoff metrics such as delay, bandwidth, jitter, packet loss etc are included

Fig 4.1 Flow Chart of variance based algorithm

Variance = ‘(x-”)’^2/N , where x is any metrics such as delay, bandwidth, jitter etc and ” is its mean of a set of samples of the particular parameters. N is set of samples.

In this algorithm, whenever the signal strength of a mobile terminal drops below threshold ,there is request from mobile terminal for handoff to the network which is accessible. The algorithm checks whether the visitor network available or not, if visitor network is available ,it will broadcast required parameters such as packet delay, jitter, packet loss and bandwidth etc. The variance of the broadcasted parameters are calculated based on the number of samples received for each parameter. Then, the candidate network (Visitor network) having most of the minimum variance of the parameter is selected.

In this case, variance of delay, jitter, packet loss and bandwidth are considered for the set of 100 samples received. Here, variance of packet delay is calculated as:

”_d^2 = ‘(D-”_d)’^2/N

Where, ”d is the variance of the packet delay parameter, D is the packet delay at that instant ,”d is the mean of the packet delay values received and N is the total number of samples for packet delay parameters(which is 100 in this case).

Similarly, variance of bandwidth is calculated as:

”_B^2 = ‘(B-”_B)’^2/N

Where, ”B is the variance of the Bandwidth parameter, B is the Bandwidth at that instant ,”B is the mean of the bandwidth values received and N is the total number of samples for bandwidth parameters(which is 100 in this case).

In the same way, variance of Jitter is calculated as:

”_J^2 = ‘(J-”_J)’^2/N

Where, ”J is the variance of the Jitter parameter, J is the Jitter at that instant ,”J is the mean of the jitter values received and N is the total number of samples for jitter parameters(which is 100 in this case).

In the same way, variance of Packet loss is calculated as:

”_P^2 = ‘(P-”_P)’^2/N

Where, ”P is the variance of the packet loss parameter, P is the packet loss at that instant ,”P is the mean of the packet loss values received and N is the total number of samples for packet loss parameters(which is 100 in this case).

Out of these variance ”_d^2 , ”_B^2, ”_J^2, ”_P^2, the candidate network most of them with minimum values will be selected.

Score” =’arg’_(i=1)^MMAX(min ”_d^2 , ”_B^2, ”_J^2, ”_P^2)

The candidate network which satisfies above equation is selected.

Where M is the number of candidate network.

4.2 SNR based vertical handoff algorithm

Proposed algorithm is an SNR based algorithm which calculates the value of SNR of parameters such as delay, jitter, bandwidth and packet loss for various networks such as UMTS, WLAN, Wimax with the network with maximum SNR being selected. In our proposed algorithm, handoff metrics such as delay, bandwidth, jitter, packet loss, etc are included

Fig 4.2 Flow Chart of SNR based algorithm

In this algorithm, the score of each candidate network is calculated as follows:

Score= 10 Log10 [mean of sum squares of reciprocal of measured Parameters such as Jitter, Packet delay, bandwidth]

In this algorithm, whenever there signal strength of a mobile terminal drops below threshold ,there is request from mobile terminal for handoff to the network which is accessible. The algorithm checks whether the visitor network available or not, if visitor network is available ,it will broadcast required parameters such as packet delay, jitter, packet loss and bandwidth etc. Otherwise mobile terminal will stay back in the current network i.e. handoff is dropped. If the visitor network is available for the given mobile terminal, the score of each candidate network is calculated as follows:

Score= ‘arg’_(i=1)^MMAX {10 Log10 [ [(1/D)2+(1/B)2 +(1/J)2+(1/P)2] /4]}

Where, D is the packet delay, B is the bandwidth, J is the jitter, P is the packet loss. M is number of network.

The candidate network having maximum score is selected and the mobile terminal sends request to the base station of the candidate network. Mobile terminal gets register to the base station of the candidate network having maximum score. Mobile terminal gets switched to the network having maximum score as per the formula mentioned above.

4.3 Proposed selective vertical handoff algorithm using ns-2

Proposed algorithm is selective algorithm which selects the best MADM algorithm viz SAW, MEW, GRA, TOPSIS etc for vertical handoff decision as per traffic. In the proposed selective algorithm, handoff metrics such as delay, bandwidth, cost, BER (bit error rate), trust level etc are included. The weight for attribute in our experiment is taken as 0.2 each. However, weights for delay, bandwidth, BER may be taken 0.25 and for cost 0.15, trust level 0.1, respectively.

Fig 4.3 Proposed selective algorithm

The best algorithm with maximum score is selected for taking vertical handoff decision.

4.4 Proposed ANN based vertical handoff algorithm

In this method, neural network is used for handoff between WLAN and Cellular network. Here, two parameters are taken into consideration i.e. RSS and Bandwidth as an input for neural network. The RSS samples for training neural network for both WLAN & cellular networks are -60dBm,-70 dBm,-80 dBm,-90 dBm. Similarly, Bandwidth samples for WLAN are 54, 30,10,1 Mbps. And Bandwidth samples for Cellular network are 14.4, 9.6, 4.5,2 Kbps. By using combination of RSS & Bandwidth parameters, we could make 256 samples of input for ANN. 256 samples of output samples for vertical handoff decision are also fed to ANN.

Using Levenberg-Marquardt method for ANN, 180 samples are used for training, 38 samples for validation and 38 samples for testing. Based on ANN developed system, it could take vertical handoff decision from cellular to WLAN and vice-versa. As Levenberg-Marquardt method is fast, training period will be much reduced.

The performance for the method is measured in terms of mean squared error (mse), which is

3.29 x 10-16 .Using the method, ping pong effect and training time are reduced.

The Levenberg marquardt algorithm was developed by Kenneth Levenberg and Donald marquardt. This algorithm provides a numerical solution to the problem of minimizing a non-linear function. It is fast and has stable convergence.

The Levenberg marquardt algorithm blends the steepest descent method and Gauss Newton algorithm. At the same time it inherits the speed advantage of the Gauss-Newton algorithm and the stability of the steepest descent method.

Chapter-5

Results and discussion

5.1 Comparison results of variance based algorithm with other MADM algorithm

In the proposed algorithm, 10 iterations are taken i.e. handoffs are taken and the network selected in each iteration is also shown in the figure below. Here, 1,2,3,4,5,6 represent UMTS1,UMTS2,WLAN1,WLAN2,Wimax1,Wimax2, respectively

In the simulation, comparative analysis of various algorithms such as SAW, TOPSIS, MEW, GRA, variance-based algorithms is given. Various graphs have been plotted on

1. Packet Delay Vs Number of Handoffs 2. Jitter Vs Number of Handoffs.

Fig 5.1 Network selection by the various algorithms

Here, the yellow line indicates the network selection of variance-based algorithm. Results for 10 iterations are also shown. The networks selected using the proposed variance based algorithms are as follow: UMTS2, Wimax1, UMTS1, UMTS2, UMTS1, UMTS1, UMTS1, Wimax1, Wimax1, UMTS1. These results are obtained for voice connections.

Fig 5.2 Packet delay vs number of handoffs for various algorithms

This graph shows Packet delay versus Number of handovers i.e. iterations. It is observed that Packet delay is minimum at variance based algorithm compared to SAW, TOPSIS, MEW, GRA algorithms. Which means that for voice connections, the proposed algorithm performs better as far as packet delay is concerned.

Packet delay for various algorithms for each handoff instance is given tabular form.

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

26.49 ms 10.6 ms 10.04 ms 26.78 ms 27.78 ms 27.41 ms 10.21 ms 9.86 ms 13.74 ms 2.74 ms

Table 5.1 Packet delay vs Handoff using SAW algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

0.265 ms 0.845 ms 0.475 0.268 ms 0.278 ms 0.445 ms 0.347 ms 0.986 ms 0.848 ms 0.274 ms

Table 5.2 Packet delay vs Handoff using variance based algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

13.91 ms 8.45 ms 9.27 ms 14.21 ms 14.61 ms 13.85 ms 7.71 ms 9.86 ms 13.52 ms 13.64 ms

Table 5.3 Packet delay vs Handoff using GRA algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

9.35 ms 13.46 ms 10.04 ms 14.21 ms 14.61 ms 13.85 ms 11.64 ms 6.73 ms 7.77 ms 7.57 ms

Table 5.4 Packet delay vs Handoff using TOPSIS algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

13.91 ms 10.6 ms 10.04 ms 2.67 ms 2.78 ms 2.71 ms 10.21 ms 9.86 ms 13.45 ms 2.74 ms

Table 5.5 Packet delay vs Handoff using MEW algorithm

Fig 5.3 Jitter vs number of handoffs for various algorithms

In this graph, Jitter versus Number of handovers i.e. iterations are given. It is observed that Jitter is minimum at variance based algorithm compared to SAW, TOPSIS, MEW, GRA algorithms. From which it can be inferred that for voice connections, the proposed algorithm performs better as far as Jitter is concerned.

Jitter for various algorithms for each handoff instance is given tabular form.

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

0.791 ms 1.17 ms 1.877 ms 0.632 ms 0.582 ms 0.559 ms 1.042 ms 0.383 ms 1.04 ms 0.8744 ms

Table 5.6 Jitter vs Handoff using SAW algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

0.007 ms 0.004 ms 0.006 ms 0.006 ms 0.005 ms 0.008 ms 0.006 ms 0.003 ms 0.006 ms 0.008 ms

Table 5.7 Jitter vs Handoff using Variance based algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

2.22 ms 1.489 ms 1.877 ms 1.353 ms 1.449 ms 1.963 ms 1.402 ms 9.69 ms 4.25 ms 8.41 ms

Table 5.8 Jitter vs Handoff using TOPSIS algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

1.503 ms 0.482 ms 0.36 ms 1.353 ms 1.449 ms 1.963 ms 0.597 ms 0.383 ms 1.189 ms 1.667 ms

Table 5.9 Jitter vs Handoff using GRA algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

1.503 ms 1.17 ms 1.877 ms 0.632 ms 0.582 ms 0.559 ms 1.042 ms 0.383 ms 1.402 ms 0.874 ms

Table 5.10 Jitter vs Handoff using MEW algorithm

Fig 5.4 Bandwidth vs Number of handoffs for variance based algorithms

In this graph, Bandwidth versus Number of handovers are Plotted .Here, the Performance for 3-7 iteration is quite poor as far as bandwidth is concerned for the proposed variance based algorithm.

From all the above three graphs, it can be observed that the proposed algorithm is best suited for voice connections. It is quite evident from above discussion that, the proposed algorithm has the lowest packet delay than MEW, SAW, GRA, TOPSIS algorithms. Moreover, it is observed that the proposed algorithm offers least jitter than any other above mentioned algorithms. These two observations make the proposed algorithm best suited for voice connections. However, the observations obtained are from the simulation results. Hence, it is recommended that for practical scenarios the proposed algorithm can be utilized to have better voice communication.

5.2 Comparison results of SNR based algorithm with other MADM algorithms

In this algorithm, the score of each candidate network is calculated as follows:

Score= 10 Log10 [mean of sum squares of reciprocal of measured Parameters such as Jitter, Packet delay, bandwidth]

In proposed algorithm, 10 iterations are taken i.e. handoffs are taken and the network selected in each iteration is also shown in the figure below. Here, 1,2,3,4,5,6 represent UMTS1,UMTS2,WLAN1,WLAN2,Wimax1,Wimax2, respectively

The simulation is done for data connections which means 70% weight age is given to bandwidth. For the rest of parameters, the weights are equally distributed.

In the simulation, comparative analysis of various algorithms such as SAW, TOPSIS, MEW, GRA, Variance based algorithms is given. Various graphs have been plotted on 1.Packet Delay Vs Number of Handoffs 2. Jitter Vs Number of Handoffs.

Fig 5.5 Packet delay vs number of handoffs for various algorithms

In this graph, packet delay versus Number of handovers i.e. iterations are given. It is observed that packet delay is quite moderate for SNR based algorithm compared to SAW, TOPSIS, MEW,GRA algorithms. This shows that for data connections, the proposed algorithm packet delay cannot be that important. Packet delay for various algorithms for each handoff instance is given in tabular form.

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

37.23 ms 106.6 ms 100.49ms 34.23 ms 38.04 ms 27.41 ms 11.64 ms 9.86 ms 8.48 ms 2.74 ms

Table 5.11 Packet delay vs Handoff using SAW algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

80.01 ms 8.451 ms 9.27ms 8.12 ms 6.8 ms 7.8 ms 7.71 ms 9.86 ms 8.48 ms 2.87 ms

Table 5.12 Packet delay vs Handoff using SNR algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

139.144 ms 8.451 ms 9.27ms 142.16 ms 146.11 ms 138.54 ms 7.71 ms 9.86 ms 135.21 ms 136.47 ms

Table 5.13 Packet delay vs Handoff using GRA algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

3.72 ms 8.451 ms 9.27ms 8.12 ms 146.11 ms 7.81 ms 7.71 ms 9.86 ms 8.48ms 8.78 ms

Table 5.14 Packet delay vs Handoff using TOPSIS algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

3.72 ms 106.6 ms 100.49ms 3.42 ms 3.8 ms 2.74 ms 116.44 ms 9.86 ms 8.48ms 2.74 ms

Table 5.15 Packet delay vs Handoff using MEW algorithm

Fig 5.6 Jitter vs number of handoffs for various algorithms

This graph shows Jitter versus number of handovers i.e. iterations. It is observed that Jitter is minimum for SNR based algorithm compared to SAW, TOPSIS, MEW, GRA algorithms. Indicating that for data connections, the proposed algorithm performs better as far as Jitter is concerned. Pink line is used for the proposed algorithm. Jitter for various algorithms for each handoff instance is given in tabular form.

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

5.81 ms 11.7 ms 18.77 ms 7.64 ms 6.59 ms 5.59 ms 14.02 ms 3.83 ms 6.46 ms 8.74 ms

Table 5.16 Jitter vs Handoff using SAW algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

3.98 ms 4.82 ms 3.6 ms 6 ms 4.8 ms 5.08 ms 5.97 ms 3.83 ms 6.46 ms 8.19 ms

Table 5.17 Jitter vs Handoff using SNR algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

5.81 ms 4.82 ms 3.6 ms 6 ms 14.49 ms 5.08 ms 5.97 ms 3.83 ms 6.46 ms 7.94 ms

Table 5.18 Jitter vs Handoff using TOPSIS algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

15.03 ms 4.82 ms 3.6 ms 13.53 ms 14.49 ms 19.63 ms 5.97 ms 3.83 ms 11.89 ms 16.67 ms

Table 5.19 Jitter vs Handoff using GRA algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

5.81 ms 11.7 ms 18.77 ms 7.64 ms 6.59 ms 5.59 ms 5.59 ms 14.02 ms 3.83 ms 8.74 ms

Table 5.20 Jitter vs Handoff using MEW algorithm

Fig 5.7 Bandwidth vs number of handoffs for various algorithms

In this graph, Bandwidth versus Number of handovers i.e. iterations are given. It is observed that Bandwidth is maximum for SNR based algorithm compared to SAW, TOPSIS, MEW, GRA algorithms. This makes it clear that for data connections the proposed algorithm performs better as far as Bandwidth is concerned. Pink line is for the proposed algorithm which shows bandwidth is highest for proposed algorithm. Bandwidth for various algorithms for each handoff instance is given in tabular form.

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

1.64 Mbps 9.54 Mbps 53.84 Mbps 1.83 Mbps 1.42 Mbps 1.42 Mbps 7.62 Mbps 58.33 Mbps 59.29 Mbps 1.93 Mbps

Table 5.21 Bandwidth vs Handoff using SAW algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

24.99 Mbps 40.34 Mbps 56.08 Mbps 48.84 Mbps 29.58 Mbps 45.64 Mbps 25.6 Mbps 58.33 Mbps 59.29 Mbps 0.7 Mbps

Table 5.22 Bandwidth vs Handoff using SNR algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

1.64 Mbps 40.34 Mbps 56.08 Mbps 48.84 Mbps 35.57 Mbps 45.64 Mbps 25.6 Mbps 58.33 Mbps 59.29 Mbps 51.98 Mbps

Table 5.23 Bandwidth vs Handoff using TOPSIS algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

34.45 Mbps 40.34 Mbps 56.08 Mbps 12.88 Mbps 35.57 Mbps 33.06 Mbps 25.6 Mbps 58.33 Mbps 2.33 Mbps 23.31 Mbps

Table 5.24 Bandwidth vs Handoff using GRA algorithm

Handoff1 Handoff2 Handoff3 Handoff4 Handoff5 Handoff6 Handoff7 Handoff8 Handoff9 Handoff10

1.64 Mbps 9.54 Mbps 53.84 Mbps 1.83 Mbps 1.42 Mbps 1.42 Mbps 7.62 Mbps 58.33 Mbps 59.29 Mbps 1.93 Mbps

Table 5.25 Bandwidth vs Handoff using MEW algorithm

From all the three graphs and explanation, we can arrive at a conclusion that the proposed algorithm is best suited for data connections. It is quite evident from above discussion that, the proposed algorithm has the moderate packet delay than MEW, SAW, GRA and TOPSIS algorithms. Besides, it is observed that the proposed algorithm offers less jitter compared to other algorithms. Furthermore, it is also observed that the proposed algorithm offers higher bandwidth than any other algorithms. These three observations make the proposed algorithm best suited for data connections. However, the observations obtained are from the simulation results. Hence, it is recommended that for practical scenarios the proposed algorithm can be utilized to have better voice communication.

Performance of Vertical handover algorithm using Artificial Neural Network

In this method neural network is used for handoff between WLAN and Cellular network. Here, two parameters are taken into consideration i.e. RSS and Bandwidth as inputs for neural network. The RSS samples for training neural network for both WLAN & cellular networks are -60dBm,-70 dBm,-80 dBm,-90 dBm. Similarly, Bandwidth samples for WLAN are 54, 30,10,1 Mbps. And Bandwidth samples for Cellular network are 14.4, 9.6, 4.5,2 Kbps. By using combination of RSS & Bandwidth parameters, we could make 256 samples of input for ANN. 256 samples of output samples for vertical handoff decision are also fed to ANN.

Using Levenberg-Marquardt method for ANN, 180 samples are used for training, 38 samples for validation and 38 samples for testing. Based on ANN developed system, it could take vertical handoff decision from cellular to WLAN and vice-versa. Owing to the fast speed of L.M method, training period will be much reduced.

Fig 5.8 Peformance for Proposed method showing mse=3.29×10-16 at epoch 746

The performance for the method is measured in terms of mean squared error (mse), which is 3.29 x 10-16

Chapter-6

Conclusion

The research came to the following conclusion

The bandwidth is maximum for SNR based algorithm compared to SAW, TOPSIS, MEW, GRA algorithms. indicating that for data connections, the proposed algorithm performs better as far as bandwidth is concerned, Bandwidth being the highest for the proposed algorithm.

Secondly it is also observed that Jitter is the lowest for SNR based algorithm compared to SAW, TOPSIS, MEW, GRA algorithms. Which means that for data connections, the proposed algorithm performs better as far as Jitter is concerned.

Thirdly Packet delay is quite moderate for SNR based algorithm compared to SAW, TOPSIS, MEW, GRA algorithms. This shows that for data connections, the proposed algorithm packet delay cannot be that important.

Packet delay is minimum for variance based algorithm compared to SAW,TOPSIS, MEW,GRA algorithms. This shows that for voice connections the proposed algorithm performs better as far as packet delay is concerned.

It can be seen that Jitter is minimum for variance based algorithm compared to SAW, TOPSIS, MEW, GRA algorithms. This shows that for voice connections, the proposed algorithm performs better as far as Jitter is concerned.

The Performance for variance based algorithm is quite poor as far as bandwidth is concerned. From all the three observations and explanation, it can be concluded that the proposed variance based algorithm is best suited for voice connections.

Future Scope

Future work can be carried out to consider other types of connections such as a) Conversational, b) Streaming, and c) Interactive. Besides this, background and study of the impact of weights on each of them can be carried out. The impact of the weight assignment may decide the performance of multiple attribute decision making algorithms. Vertical handoff execution procedure may be considered too. Since the observations in this thesis are obtained from simulation results, it is recommended that the proposed algorithm can be utilized to have better voice communication for practical scenarios.

References

[1] T. S. Rappaport, Wireless Communications: Principles and Practice.Prentice Hall, July 1999.

[2] Enrique Stevens-Navarro and Vincent W.S. Wong ‘Comparison between Vertical Handoff Decision Algorithms for Heterogeneous Wireless Networks’,2006,IEEE pp 947-951

[3] J.D Martinez- Morales, V.P.rico, E.Steven, ‘Performance comparison between MADM algorithms for vertical handoff in 4G networks’, 7th International Conference on Electrical Engineering Computing Science and Automatic Control (CCE), 2010, pp. 309-314.

[4] Xiaohuan Yan,.Ahmet S” ekercio”glu, Sathyanarayan ‘A Survey of vertical decision algorithms in fourth generation heterogeneous networks’Elsevier,2010, pp.1848-1863

[5] A. H. Zahran and B. Liang, ‘ALIVE-HO: Adaptive lifetime vertical handoff for heterogeneous wireless networks,’ Technical Report, University of Toronto

[6] S. Mohanty and I. F. Akyildiz, ‘A cross-layer (Layer 2 + 3) handover management protocol for next-generation wireless systems,’ IEEE Trans. Mobile Computing, vol. 5, pp. 1347’1360, Oct. 2006.

[7] K. Yang, I. Gondal, B. Qiu, and L. S. Dooley. Combined SINR based vertical handoff

algorithm for next generation heterogeneous wireless networks. In Proceedings of the

2007 IEEE Global Telecommunications Conference (GLOBECOM’07), pages 4483-4487,

Washington, DC, USA, November 2007.

[8] C. Chi, X. Cai, R. Hao, and F. Liu. Modeling and analysis of handover algorithms. In

Proceedings of the 2007 IEEE Global Telecommunications Conference (GLOBECOM’07),

pages 4473-4477, Washington, DC, USA, November 2007.

[9]A. Hasswa, N. Nasser, and H. Hassanein. Tramcar: A context-aware cross-layer archi-

tecture for next generation heterogeneous wireless networks. In Proceedings of the 2006

IEEE International Conference on Communications (ICC’06), pages 240{245, Istanbul,

Turkey, June 2006.

[10] R. Tawil, G. Pujolle, and O. Salazar. A vertical handoff decision scheme in heterogeneous wireless systems. In Proceedings of the 67th Vehicular Technology Conference (VTC’08- Spring), pages 2626{2630, Marina Bay, Singapore, April 2008.

[11] N. Nasser, S. Guizani, and E. Al-Masri. Middleware vertical handoff manager: A neural

network-based solution. In Proceedings of the 2007 IEEE International Conference on

Communications (ICC’07), pages 5671-5676, Glasgow, Scotland, June 2007.

[12] K. Pahlavan, P. Krishnamurthy, A. Hatami, M. Ylianttila, J. P. Makela, R. Pichna, and

J. Vallstron. Handoff in hybrid mobile data networks. IEEE Personal Communications,

7(2):34-47, 2000.

[13] Xiaohuan Yan,N,Nallasamy Mani, and Y.Ahme t Sekercio”glu, ‘ A Traveling Distance Prediction Based Method to minimize Unnecessary Handovers from Cellular Networks to WLANs,’ IEEE communication letters, vol. 12, pp. 14’16. 2008.

[14] Ahmed H. Zahran and Ben Liang ‘Performance Evaluation Framework for Vertical Handoff Algorithms in Heterogeneous Networks’,in :Proceedings of the 2005 IEEE International Conference on Communications(ICC05),Seoul,Korea,May2005. pp.173-178

[15] QingHe,Guang Chen,Le Zhang ‘A Vertical Handoff Decision Algorithm Based on Fuzzy Control in WiMax and TD-SCDMA Heterogeneous Wireless Networks’,2010,IEEE

Author’s Publication

[1] Abhijit Bijwe, Dr.C.G.Dethe, ‘ RSS based Vertical Handoff algorithms for Heterogeneous wireless networks – A Review’ IJACSA, Wireless & mobile networks, Special issue no.2,pp.62-67 , Jan2011

[2] Abhijit Bijwe, Dr.C.G.Dethe ‘Analysis of Vertical handoff parameters using novel algorithm’ IJISET, Vol 2 issue 4 ISSN 2348-7968,April 2015 pp-853-858

[3] Abhijit Bijwe, Dr.C.G.Dethe ‘Performance Analysis of Heterogeneous network using Vertical handoff algorithms for Wireless Communication IJRTET,ACEE got selected under POSTER category, 2014

[4] Abhijit Bijwe,Dr.C.G.Dethe, ‘Vertical Handoff algorithms using Neural networks’ Proc. of the Second Intl. Conf. on Advances in Computer, Electronics and Electrical Engineering — CEEE 2013 ,pp 39-42 , April2013

[5] Abhijit Bijwe,Dr.C.G.Dethe ‘Performance Analysis of Heterogeneous network using vertical Handoff algorithms for wireless communication’ Proc. of the Intl. Conf. Of MICTM School of engineering and information Technology at Manipal University UAE,Dubai, 25-26th March 2015 Vol 4 pp-13

[6] Abhijit Bijwe,Dr.C.G.Dethe ‘Performance Analysis of Vertical handoff metrics

using Variance based algorithms’ IJARCCE, August 2015 Vol 4 issue 8 pp-487-490

**...(download the rest of the essay above)**