Content uploaded by Ying Wang
Author content
All content in this area was uploaded by Ying Wang on Nov 13, 2014
Content may be subject to copyright.
Vertical Handover Decision in An Enhanced Media
Independent Handover Framework
Wang Ying1,2, Yuan Jun2, Zhou Yun2,LiGen
2, Zhang Ping1,2
Key Laboratory of Universal Wireless Communications, Ministry of Education1
Wireless Technology Innovation Institutes2, Beijing University of Posts and Telecommunications
P.O. Box 92, BUPT, Beijing, China
Email: wangying@bupt.edu.cn
Abstract—Vertical handover decision making is one of key
problems in heterogeneous network environment. In IEEE 802.21
standard, a Media Independent Handover (MIH) framework is
presented to facilitate handover with measurements and triggers
from link layers. However, vertical handover decision making can
benefit from the information more than link layers. In this paper,
an Enhanced Media Independent Handover (EMIH) framework
is proposed by integrating more information from application
layers and user context information. Given such information,
the issue becomes how to select a favorite network. In this
paper, two novel weighted Markov chain (WMC) approaches
based on rank aggregation are proposed, in which a favorite
network is selected as top one of rank aggregation result fused
from multiple ranking lists based on decision factors. The
proposed approaches can easily integrate a priori knowledge
and/or human experiences into vertical handover. Simulation
results demonstrate the effectiveness of the proposed approaches.
Index Terms—Enhanced Media Independent Handover, het-
erogeneous networks, vertical handover, rank aggregation;
I. INTRODUCTION
Integrated all-IP network has great potential to provide
better services to the subscribers. However big differences
between traditional and heterogeneous network environment
make traditional handovers no longer satisfy the requirements
in the new environment. Vertical handovers is therefore nec-
essary in heterogeneous networks environment. A new speci-
fication namely IEEE 802.21 (Media Independent Handover)
[1], is emerging to provide link intelligence and other related
network information to upper layers. Reasonable handover
decision is then expected to be obtained and user experience of
mobile devices is intended to be enhanced. Two major issues
in Media Independent Handover (MIH) are as follows:
•Collection mechanism for information in both net-
work side and mobile node
A comprehensive discussion is given to collection mech-
anism in MIH with several limitations as follows:
–Measurement and trigger mechanism is only working
on the link layers of terminals for handover decision
making;
–A timely update of information is not supported by
MIH.
In this paper, a new framework namely enhanced me-
dia independent handover (EMIH) is presented for the
collection mechanism. More decision factors and related
trigger events are suggested in upper layer. Furthermore,
different handover types are specifically studied and sup-
ported by EMIH, such as mobile controlled and network
controlled.
•Vertical handover decision
Many efforts have been focused on the topic of vertical
handover, which is roughly categorized into policy based
[2], fuzzy logic based [3], [4], [5], and multiple attribute
decision making (MADM) based approach [6], [7], [8].
A certain performance can be achieved by these ap-
proaches. However, some realistic problems are still not
discussed explicitly. In general, vertical handover can
benefits much from a priori knowledge and /or human ex-
periences. Two WMC approaches are therefore proposed,
in which a favorite network is selected as top of rank
aggregation result fused from multiple ranking lists based
on decision factors. The new approaches easily integrate a
priori knowledge and/or human experiences into vertical
handover.
The remainder of this paper is organized as follows. In
section II-A, the framework of EMIH is presented. In section
II-B, a new set of decision factors and related trigger events
are proposed. In section III-A, two kinds of WMC based
approaches are proposed. In section III-B, simulations are
performed to evaluate the effectiveness of our proposed WMC
based approaches. In section IV, an implementation of whole
vertical handover is demonstrated. Finally conclusions and
discussions are given in section V.
II. A FRAMEWORK OF EMIH
In this section, a new framework of enhanced media in-
dependent handover is presented and a new set of decision
factors and related trigger events are proposed.
A. Enhanced Media Independent Handover Framework
An enhanced media independent handover framework is
proposed to improve the performance of mobility management
between heterogeneous networks, which is shown as Fig. 1.
The motivation is to make full use of available information
in both client side and network side to optimize handovers.
New function entities and modules are introduced to provide
link layer, application layer, user and network information
1525-3511/08/$25.00 ©2008 IEEE
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2008 proceedings.
to mobility decision engine. EMIH deploys comprehensive
trigger event criteria and flexibly collects the static and dy-
namic information available at mobile node (MN) and within
the network infrastructure, which will be used to optimize
the handover decision making in the proposed framework.
It should be emphasized that overall mobility management
architecture possibly includes Mobile IP infrastructure (client,
HA and so on)[9] or any other mobility schemes. EMIH
benefits from the application dependent information and user-
aware information on mobility. Fig. 1 illustrates the key
entities both in client side and network side.
•Client Side:
–EMIHF (EMIH Function): EMIHF is a logical entity
to provide link layer intelligence and offer a unified
interface between different access schemes and upper
layer applications.
–CAM (Context-Aware Module): CAM identifies in-
formation of MN, generates trigger events including
Application QoS Change and User Aware Change,
transfers events and related information to HCM.
–HCM (Handover Control Module): HCM has capa-
bilities to support MN controlled handover. There are
two sub-modules in HCM, namely trigger FE and
handover FE. Once trigger FE decides to initiate a
handover, it notifies handover FE and handover FE
selects a favorite network.
•Network Side:
–Access Network (AN): EMIHF, HCM also exist
in AN. HCM has capabilities to support network
controlled handover. trigger FE receives the trigger
events and handover FE controls handover.
–MIIS Server: MIIS Server is a function entity in-
cluding EMIHF and Information Service module.
Network related information collected in MIIS can
be accessed by EMIHF in other entities.
–CAS: CAS identifies network context, generates trig-
ger events and transmits these events to subscribers
(e.g. HCM in AN) through EMIHF.
–CEMIH (Control EMIH): CEMIH belongs to a cen-
tralized control entity and provides some controlling
functions. Several functions in CEMIH includes 1)
collecting trigger events; 2) initiating a handover; 3)
controlling handover signaling to pass core network;
4) selecting a target network.
In summary, all logical entities communicate with each other
through EMIHF, which is implemented in either client side or
network side. CAM or CAS identifies the useful information
(e.g. application layer information, user context or network
context). New trigger events are generated and then transmitted
to HCM through EMIHF. HCM makes used of information
of Lower (L2/L1) layer and higher layers from client side or
network side. A reasonable decision can be obtained thereafter.
B. Decision Factors and Related Trigger Events
In EMIH, New decision factors and related trigger events
are defined more than link layer triggers and generated by
HCM
EMIHF
Information
Service module
EMIHF
CAM
EMIHF
EMIHF
HCM
EMIHF
HCM
EMIHF
HCM
EMIHF
CAM HCM
CEMIH MIIS Server CAS
AN1 AN2 AN3
MN
Client Side
Network Side
Fig. 1. EMIH Framework
sources such as application layer as follows.
•Link information
Event service for link information is the same as the
definition in MIH. Categories of events include MAC
and PHY State Change events, Link Parameter events,
Predictive events, Link Synchronous events and Link
Transmission events.
•Application QoS
Trigger event for application QoS is Applica-
tion QoS Change and indicates the change of user traffic
and the corresponding QoS parameters. Traffic class
indicates QoS class of service including conversational
class, streaming class, interactive class and background
class [10], which have different requirements of
bandwidth, packet loss, delay and jitter.
•User context
Trigger event for user context is User Aware Change
and indicates the change of user context information.
User context is composed of location information, mobile
mode, user preference and user instruction, in which a
reasonable handover is favorable. In terms of location in-
formation, there are various environments such as urban,
suburban or rural regions. Mobile mode consists of indoor
mode, high speed mobile or nomadic mode, in which
handover is triggered by the change of mobile speed and
directions. User preference indicates cost and/or energy
preferable by users and user instruction means the favorite
network for a specific user.
•Network context
Trigger event for network context is Net-
work Context Change and indicates the change of
network context. Network context is made up of load
information, available resource, throughput and security
level. In particular, security level is one of key decision
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2008 proceedings.
factors and can not be replaced by others. It should be
emphasized that low security level can not be accepted
by most users even when the network performs well.
In this paper, seven decision factors are used for vertical
handover, which are listed as follows:
•Total bandwidth (TBW): TBW indicates how much
bandwidth is available for a candidate network.
•Allowed bandwidth (ABW): ABW indicates the band-
width allowed by the candidate network for a single user.
•Cost per byte (C): C means relative transport cost of the
operator for a particular access network.
•Load (Ld): Ld represents the ratio of allocated bandwidth
to the total bandwidth.
•Delay (D): D represents the average packet delay within
the network.
•Jitter (J): J measures the average delay variations within
the network. A large J could result in packet reordering
or dropping of real-time packets at the receiver.
•Packet loss (L): L measures the average packet loss rate
within the nework.
III. RANK AGGREGATION BASED HANDOVER DECISION
A. Weighted Markov Chain based Approach
Vertical handover with multiple decision factors can be
formulated as a rank aggregation problem, in which a ”better”
ranking can be derived by combining ranking results of the
different decision factors. Recently, more and more efforts are
focused on the topic of rank aggregation because this topic is
one of the key issues in web search area.
In [11], three methods namely linear combination method,
Borda count and Markov chain (MC) method are reviewed
and compared with web search application. Among three
methods, MC method is preferable in integrating application
dependent heuristics, which make it very attractive to vertical
handover with multiple decision factors. MC method begins
by constructing a Markov chain transition matrix on given
ranking lists. A priori knowledge and/or user experience can
be naturally involved with particular definition of element
in Markov chain transition matrix and particular weight for
ranking lists of decision factors. Stationary probability distri-
bution is then derived and used to sort candidate networks. The
favorite network is finally selected as the candidate network
with the largest value of stationary probability. In this sec-
tion, two kinds of weighting Markov chains methods WMC1
and WMC2 are proposed with difference on construction of
Markov chain transition matrix as follows.
Consider a candidate network set P={p1,...,p
N}and a
decision factor set Q={q1,...,q
M}where Nis the number
of candidate networks and Mis the number of decision
factors. For decision factor q, a ranking list is obtained as
an ordering of P, i.e. τq=pq
1≥pq
2≥ ··· ≥ pq
Nwhere ”≥
” represents some ordering relation on Psub. Also, let τq(p)
denote the position or rank of pin τq.
With above definitions, WMC1 method is described as
follows:
1) Normalization of decision factor weight:
Every decision factor is given a normalized weight in
which the weight of decision factor qis denoted by wq
with constraint as q∈Qwq=1
2) Construction of weighted Markov chains transition
matrix MC:
a) Initialize a N×Ntransition probability matrix
MC ={mcij }with all elements equal to zero, in
which mcij represents transition probability from
pito pj.
b) For each τq,q∈Q,MC is updated as follows:
i) For each mcij in MC, update
mcij =mcij +wq
τq(pi)(1)
if pi,p
j∈Pand τq(pi)≥τq(pj)
ii) Repeat the above step until all τq,q∈Qare
examined.
3) Computation of stationary distribution (row) vector
SD:
SD =SD ×MC (2)
where SD ={sd1,...,sd
N}and sdnis the element
for network pn.
4) Selection of favorite network pγ:
γ= arg max
nsdn(3)
WMC2 method is similar to WMC1 with only difference
in step 2(b) as follows:
b) For each τq,q∈Q,MC is updated as follows:
i) For each mcij in MC, Update
mcij =mcij +wq
N(4)
if pi,p
j∈Pand τq(pi)>τ
q(pj)
mcij =mcij +N−τq(pi)+1
Nwq(5)
if pi,p
j∈Pand τq(pi)=τq(pj)
ii) Repeat the above step until all τq,q∈Qare
examined.
According to [12], two preference weighting methods are
presented in which weights assignment is based on either
service type or user subscription level. In this paper, main
discussion is focussed on the approaches on vertical handover.
So a uniform weighting method, in which all weights are equal
to 1/M , is adopted to simplify the comparative experiments
between proposed approaches and traditional approach.
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2008 proceedings.
U1W1
W2
U2
BS-
U2
AP1
AP2
MT1
BS-
Wi2
BS-
U1
BS-
Wi1Wi2
Wi1
MT2
MTn
Fig. 2. Simulation Scenario
B. Performance Evaluation
In order to compare performances of different approaches,
a multi-user scenario is designed and illustrated in Fig. 2.
The candidate network set consists of two UMTS networks
(U1,U
2), two WiMax networks (Wi
1,Wi
2)and two WLAN
networks (W1,W
2). For each pair of (Un,Wi
n),n =1,2,
the corresponding base stations BS−Un,BS−Wi
nhave the
same positions for n=1,2. Moreover the coverage radiuses
in a certain cell by UMTS, WiMAX and WLAN are set to
1000m,1000mand 100mrespectively. In the simulation,
there are four types of service consisting of VoIP and three
kinds of data services with different bandwidth requirements.
The bandwidth requirement of VoIP service is 12.2kbps and
the minimum bandwidth requirement for three kinds of data
services are 0.5Mbps,1Mbps and 2Mbps respectively. The
arrival rates of both VoIP service and three data services follow
the Poisson distribution and the arrival rate ratio between VoIP
and data services is equal to 2:1. Mean holding time follows
exponentially distribution with parameters as mean denoted
by 1/µvfor VoIP service and mean denoted by 1/µdfor data
services. In the simulation, 1/µvis set by 120 seconds and
1/µdis set by 300 seconds. In the meanwhile, the velocity of
every mobile user is set to 0.8m/s.
In the simulation, WMC1, WMC2 and TOPSIS [6] are
studied. At the beginning of simulation, the decision factor
values are initialized as table. I, which is then updated in the
following period. The simulation results are illustrated in Fig.
3, Fig. 4 and Fig. 5 respectively.
In Fig. 3, the performance on mean delay of VoIP users with
WMC1 is consistently better than TOPSIS when the arrival
rate varies from 0.8 to 2.0. In Fig. 4 on mean delay of data
users, it can be observed that there is no significant perfor-
mance difference among the WMC1, WMC2 and TOPSIS. In
Fig. 5, WMC1 and WMC2 are both better than TOPSIS.
IV. IMPLEMENTATION OF MOBILITY MANAGEMENT
The process on effective mobility management in EMIH
architecture mainly includes three steps as follows:
TAB L E I
INITIALIZED DECISION FACTOR S VALUES
D L ABW TBW CLd
(ms) (per106) (mbps) (Mbps) (price) (%)
UMTS1 35 70 0.5 20.6 0
UMTS2 30 80 0.6 20.8 0
WLAN1 100 20 111 0.1 0
WLAN2 140 18 1.5 54 0.05 0
WiMAX1 60 15 2.5 100 0.5 0
WiMAX2 70 20 3100 0.4 0
0.8 1 1.2 1.4 1.6 1.8 2
0.0425
0.043
0.0435
0.044
0.0445
0.045
0.0455
0.046
0.0465
Arrival Rate[users/sec]
Mean Delay of VoIP Users[s]
WMC1
WMC2
TOPSIS
Fig. 3. Mean Delay of VoIP Users
0.8 1 1.2 1.4 1.6 1.8 2
0.065
0.066
0.067
0.068
0.069
0.07
0.071
0.072
0.073
0.074
0.075
Arrival Rate[users/sec]
Mean Delay of Data Users[s]
WMC1
WMC2
TOPSIS
Fig. 4. Mean Delay of Data Users
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2008 proceedings.
0.8 1 1.2 1.4 1.6 1.8 2
4
4.5
5
5.5
6
6.5
7x 104
Arrival Rate[users/sec]
Total Cost
WMC1
WMC2
TOPSIS
Fig. 5. Total Cost
A. STEP 1: Obtaining the trigger events and information
related with Handover
Two mechanisms are utilized to obtain the trigger events
and the related information for EMIH users (HCM or upper
layers).
1) Registration mechanism: EMIH users specify events to
receive notifications from EMIH Function. MIH users
specify additional parameters to control the behavior of
the Event Service.
2) Query/response mechanism: EMIH users send a request
to CAM, CAS or MIIS Server. The response includes
information either in client side or in network side.
B. STEP 2: Handover decision making, network selection and
resource negotiation
Once a handover is triggered, a favorite network is selected
by WMC based rank aggregation approach. Current network
then sends handover preparation request to target network,
with the information of MN capability and context. The
MN context includes a permanent user identity and other
information, e.g. security and IP bearer parameters. The target
network will reserve resources for MN command to reduce
interruption time.
C. STEP 3: Handover execution and resource release
Mobile IP (MIP) is one of possibilities for mobility man-
agement in 4G mobile system [9]. After link layer handover,
MIP signaling will be exchanged over radio interface to
update route. Bi-casting or data forwarding mechanism may
be deployed to minimize packet loss. Finally, the resources in
source network will be released.
EMIH supports various handover types. For example, han-
dover can be controlled by MN or Network. The type of
handover can be selected when the signaling needs to be
centralized by CEMIH.
In addition, handover can be initiated either by MN or
by network. Fig. 6 illustrates the process of MANC (Mobile
assistant network control) handover, in which there is no
unified control by CEMIH and handover is initiated by MN.
Some differences may exists in different handover types. The
detailed procedure is as follows:
1) MN is associated to AN1. CAM of MN checks appli-
cation layer and user context continually;
2) CAS in network side identifies network context of AN1;
3) CAS identifies network context of AN2 simultaneously;
4) According to collected information, CAM or CAS
makes decision on generating trigger event;
5) Application QoS Change is triggered due to higher
bandwidth requirement of a new application. The related
user context is carried by this trigger and transmitted to
HCM of AN1 through EMIH Function;
6) After received trigger event, HCM queries dynamic
network information from CAS accordingly;
7) CAS responds to HCM with related information;
8) After received trigger event, HCM queries static network
information from MIIS server accordingly;
9) MIIS server responds to HCM with related information;
10) Handover is executed.
After handover is performed, maybe there are some changes
of network status (e.g. data rate, available bandwidth). There-
fore, in order to guarantee users experience, related QoS in-
formation in application layer should be adjusted accordingly.
A new command service named Application QoS Adjust is
defined in this paper. It can be used to adjust the QoS infor-
mation of application before or after handover. This command
service will be transferred through EMIHF from network side
to client side.
V. C ONCLUSION AND DISCUSSION
This paper discusses an enhanced media independent han-
dover framework and its mobility management mechanism
based on IEEE 802.21, in which new FEs and modules are pre-
sented. Comprehensive trigger criteria and handover schemes
are provided for seamless mobility management. This mech-
anism supports adaptive adjustment according to application
change, user and network information. All events follow basic
criterion in IEEE 802.21. EMIH and its mobility management
mechanism are easy to implement in IEEE 802.21. A new
rank aggregation approach is also proposed for handover
decision. Experimental results demonstrate the effectiveness
and good potential of proposed approaches. Ongoing and
future works includes 1) studying application based weighting
method and weighting fusion method with multiple classes
of weights; 2) studying more efficient and more application
dependent Markov chain generation method; 3) studying better
performance evaluation method; 4) investigating effectiveness
of incremental rank aggregation approach by considering
feedback of user and network. More results will be reported
elsewhere when they are available.
ACKNOWLEDGEMENT
This paper is supported by National Natural Science Foun-
dation of China (Project 60772112).
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2008 proceedings.
MN
CAM EMIHFHCM EMIHFHCM
AN1
EMIHFHCM
AN2
MIIS
Server CAS
measurement
1、CAM identifies application layer and user context
4、information processing
2、CAS identifies network context of AN1
3、CAS identifies network contex t of AN2
4、information processing
5、Application_QoS_Change （co ntaining related information）
6、HCM queries dynamic network informat ion from CAS accordingly（Opt.）
7、respond to HCM（containing require d information）
8、HCM queries static network inf ormtion from IIS accordingly（Opt.）
9、respond to HCM（containing required information）
handover decision, handover preparation, handover execution, handover complete
trigger
Fig. 6. MANC Handovers
REFERENCES
[1] IEEE P802.21/D05.00Draft Standard for Local and Metropolitan Area
Networks: Media Independent Handover Services, April 2007.
[2] H. J. Wang, R. H. Katz and J. Giese, “Policy-enabled handoffs across
heterogeneous wireless networks,” in Proc. WMCSA-1999, pp. 51-60,
1999.
[3] Y. Nkansah-Gyekye and J. I. Agbinya, “Vertical handoff between WWAN
and WLAN,” in Proc. ICNICONSMCL-2006, pp. 23-29, 2006.
[4] Fatma Tansu and Muhammed Salamah, “On The Vertical Handoff Deci-
sion for Wireless Overlay Networks,” in Proc. ISCN-2006, pp. 111-115,
2006.
[5] H. W. Liao, L. Tie, Z. Du, “A Vertical Handover Decision Algorithm
Based on Fuzzy Control Theory,” in Proc. IMSCCS-2006, Vol. 2, pp.
309-313, 2006.
[6] W. Zhang, “Handover Decision Using Fuzzy MADM in Heterogeneous
Networks,” in Proc. WCNC-2004, Vol. 2, pp. 653-658, 2004.
[7] Q. Y. Song, A. Jamalipour, “A Network Selection Mechanism for Next
Generation Networks,” in Proc. ICC-2005. Vol. 2, pp. 1418-1422, 2005.
[8] E. Stevens-Navarro,V. W. S. Wong, “Comparison between Vertical Hand-
off Decision Algorithms for Heterogeneous Wireless Networks,” in Proc.
VTC-2006. Vo. 2, pp. 947 - 951, 2006.
[9] D. B. Johnson, C. Perkins and J. Arkko, “Mobility Support in IPv6,” RFC
3775, June, 2004.
[10] 3GPP, “Quality of Service(QoS) Concept and Architecture,” TS
23.107(v.7.1.0), September 2007.
[11] C. Dwork, R. Kumar, M. Noar, and D. Sivakumar, “Rank aggregation
methods for the web,” in Proc. ICWWW-2001, pages 613-622, 2001.
[12] F. Bari, V. C. M. Leung, “Automated network selection in a heteroge-
neous wireless network environment,” in Proc. Network-2007, pp. 34-40,
2007.
[13] D. L. Mon, C. H. Chang and J. C. Lin, “Evaluation weapon system
using fuzzy analytic hierarchy process based on entropy weight,” FSS-
1994, Vol. 62, pp. 127-134, 1994.
This full text paper was peer reviewed at the direction of IEEE Communications Society subject matter experts for publication in the WCNC 2008 proceedings.
