You are on page 1of 5

The 14” IEEE 2003 International Symposium on Persona1,lndoorand Mobile Radio Communication Proceedings

Performance Analysis of Vertical Handover in a


UMTS-WLAN Integrated Network
Hongyan Bing, Chen He and Lingge Jiang
Department of Electronic Engineering. Shanghai Jiao Tong University
Shanghai 200030. China
{hybing, chenhe} @sjtu.edu.cn

AI~srrncf-The 4C wireless communication system is envisioned In this paper we propose an architecture integrating
to integrate cellular network (e.g. UMTS) and wireless LAN UMTS and 802.11 WLAN at SGSN, what is called tight-
(WLAN), in virtue of the individual coverage and capacity coupling architecture in some articles, in which we introduce a
complementary characteristics. This paper describes a possible
functional entity, network interwork unit (NIU), placed
architecture integrating UMTS and 802.11 WLAN a t SGSN. A
functional entity, network interwork unit (NIU), i s placed
between the SGSN and WLAN access network side to hide the
between the SGSN a n d WLAN access network side to hide the WLAN particularities. The interworking mechanism can
WLAN particularities. I n this integrated architecture, M o b i l e IP effectively combine two access networks. However, Mobile IP
based vertical handover management increases the based vertical handover protocol increases the heterogeneous
heterogeneous system signaling cost as well as handover latency, system signaling cost and produces high handover delay. So
so one major task in the handover design is to reduce the one major task in the handover design is to reduce unnecessary
unnecessary handover probability. This paper presents an handover rate and calls-dropping probability.
analytical vertical handover initiation model based on the criteria This paper presents an analytical vertical handover
o f received signal strength (RSS) and distance. T h e handover
initiation model based on received signal strength (RSS) and
performance evaluation criterion of interest is the handover
probability. T h e numerical results show that our model is i n line
distance criteria. The handover performance evaluation criteria
with the real cases when the distance between the base station of interest is handover probability and calls-dropping
(BS) o f UMTS cell and the access point (AP) of WLAN is long or probability, but only handover probability is evaluated here for
not. space limitation. The remaining part of this paper is organized
Kql~wcircl.~--heterog.eiieous networh, irttegmtion. vertical as follows. In Section 11, the interworking architecture of
/iitndoverq“ d y i c c t l niodel, perforinnnce UMTS and WLAN is described with introducing the entity
NIU. Section 111 details the vertical handover initiation
I. INTKODUC I ION analytical model based on RSS and distance criteria.
Numerical evaluation of handover probability is given i n the
The nest generation network will enable different access next section together with analyses. Finally, Some conclusions
networks to interoperate with each other to ensure global are provided in Section V.
mobility and service continuity. The deployment and
integration of both land mobile UMTS network and WLAN are
necessary thanking for the individual coverage and capacity 11. HETEROGENEOUS
NETWORK
ARCHITECTURE
complementary characteristics of the two radio technologies.
Operators with cellular network could add WLAN as an In UMTS [7], RNCs and Node Bs constitute the radio
additional service, enabling them to provide their customers access network (RAN), called UMTS RNS. Each Node B
with the broadest coverage and access possibilities for voice, constitutes a cluster of base stations and a group of Node Bs is
data and multimedia services, especially for the areas where connected to a single RNC. The packet core network (CN),
there is a high density of users. evolved from the GPRS core network, is comprised of SGSN
The main concept of the integration of UMTS and WLAN and GGSN. As for the integration, WLAN can be connected to
is to achieve a ubiquitous coverage across two radio either RNC, or SGSN, or GGSN [6].However, connecting the
technologies. The users can choose the best available WLAN at RNC requires revision of complex radio procedures
connection for the application they are using at the active time. implemented at RNC because the two radio interfaces are
This cal Is for efficient interworking mechanisms between totally different. Alternatively, connecting the WLAN at
UMTS cellular networks and WLAN. Several approaches have GGSN suffers a delay handover because the SGSN needs to
been proposed for interworking networks. The European recreate mobility state and the session (PDP) contexts. So we
Telecommunications Standards Institute (ETSI) specifies two choose the SGSN as the integration point for our integrated
generic approaches for inteiworking [I]: loose coupling and architecture. Fig. 1 shows the proposed architecture
tight coupling. The Third Generation Partnership Project configuration. To the SGSN, WLAN is regarded as one type of
(3GPP) has specified sis interworking scenarios [2]. Reference radio access network (RAN) while the UMTS RNS as a second
[3] also describes five different architectures for implementing type of RAN. This system requires less modification of
handover between GPRS and 802. I I WLAN networks. existing access networks.
Notable references to other interworking architectures are also Mobile station in Fig.1 is a multiple-mode mobile station
could be found in [4-6], etc. (MMS) supporting both UMTS and WLAN access in a

0-7803-7822-9/03/$17.0002003 IEEE.
187

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:42:23 UTC from IEEE Xplore. Restrictions apply.
The 14thIEEE 2003 International Symposium on Persona1,lndoorand Mobile Radio Communication Proceedings

based on various parameters such as signal level, the distance


between MT and BSIAP, the active user’s moving velocity and
the current network load, etc. For analysis simplicity, we
assume that the user is moving along a straightaway and the
network load is not heavy, so only signal level and distance
criteria are considered in our model. In addition, because
WLAN can provide users more bandwidth at lower cost,
handover to WLAN should be done with high priority although
cellular network coverage is available everywhere.

A. RSSCriterion .
The channel propagation model used for RSS is given by
[91[w
RSS(d)= Pt - P L ( d ) + f ( p , ~ )dBm (1)
Fig. I .The integrated architecture configuration where Pt is the transmit power and PL(d) is the path loss at
distance d and f ( p , a ) is a Gaussian random variable with
seamless fashion. The key functional entity in the system is the
network interwork unit (NIU), which is connected to a serving mean ,U and standard deviation CT modeling shadow fading. In
GPRS support node (SGSN) and WLAN access point (AP), this model, the effect of fast fading is neglected since the
assuming only one WLAN cell is considered to interwork with samples of RSS are averaged. CT depends on the environment
cellular cells here. The main function of the NIU is to provide a with values between 6-12 dB. The path loss at distance d is
standardized interface to the core network and hide the WLAN given by
particularities. So the NIU should provide the functions such as
data encapsulation, signaling transferring. With the bridging P L ( d ) = S + 10. n.log(d) dB (2)
aid of NIU, the transport UMTS signaling and data over where S is the path loss constant depends on propagation
WLANs comes true. environment and n is the path loss exponent with values
between 2-4. Fig.2 (a) and (b) illustrate the RSS on cellular link
and WLAN link respectively.
111. UMTS-WLAN VERTICAL HANDOVER
We assume that the user locates in the area of WLAN
The term vertical handover originates from overlay where is also overlaid by three UMTS cells (Cell 1, Cell 2 and
networks [SI, where different networks are mutually overlaid Cell 3), as shown in Fig.3. The total number of surrounding
offering different size of coverage. Traditional handover channels monitored by the MT during the call is three cellular
(horizontal handover) is simply migrating within the same links plus one WLAN link. If an active MT in Fig.3 is using
network and vertical handover refers to the handover across UMTS cell link, it’s obvious that RSS on Cell 1 link is
two cells from different networks. Furthermore, the generalized maximal among the surrounded three UMTS cells, i.e.
vertical handover is regarded as the handover from one
wireless access technology to another. As an example in our
work, vertical handover refers to the handover between UMTS
{ ri,(x) = mas [-r : , ( x ) , k = 11
1,2,3 (31
cell and WLAN. We are aware that the current vertical So we regard Cell I as the candidate.
handover proposal is mostly based on Mobile IP protocol, A simple situation is considered that the BS - A P direction
which is a generic technique to perform handover of IP traffic is the horizontal axis direction along which the user is moving
between IP subnets. With the help of bi-tunneling and routing with velocity v . The distance from AP to BS 1 is denoted
optimization, it solves most of triangular routing and firewall asDAP-BS. When an active MT is locating at position x, let
traversal problems and is well suited to do vertical handover - -
[6]. However, the lumping signaling across networks for r,(x) and r,(x) be the respective received signal level
vertical handover will increase the handover delay and call- averaged over the UMTS radio link and the WLAN radio link.
dropping probability. They can be described to have a lognormal distribution if
__
This section presents an analytical vertical handover model represented in decibels [ I 11, thus we have q , ( x )
-
-
and we focus on the vertical handover probability analyses in
the next section. Vertical handovers in our system fall into two
N [Y, -
(x), 0; 1 and r, (XI N [p, ( x ) , 0; I , in which P,/
and U
,,

directions: handover from UMTS cell to WLAN and handover are the mean signal levels, and and 0, are the shadowing
- -
from WLAN to UMTS cell. Two stages are included during a standard deviations of v,, ( x ) and r, ( x ) . If the initial standard
vertical handover initiation process: handover gathering deviation on each link is ao0 and owe, after an average over N
information stage and handover decision stage. In the handover
gathering information stage, various parameters used for samples, the relations of the standard deviation before and after
handover decision-making are continuously monitored by both average on each link are given by [ 1 11
the network and mobile terminal (MT). In the handover
decision stage, the handover target direction is chosen based on
the predefined criteria. Handover in both directions can be

188

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:42:23 UTC from IEEE Xplore. Restrictions apply.
The 14mIEEE 2003 International Symposium on Persona1,lndoorand Mobile Radio Communication Proceedings

z? -40 .'
?6 -60-
(0

-80 -

Distance fa cellular cell RS ( m )


-100
0
' 20 40 60 80 3
,04 Distance 10 WLAN AP ( m

(a) (b)
F i g 2 RSS on UMTS cell link and WLAN link (a): UMTS cell link; (b): WLAN link

distribution tdn - N [0,$,, ] , where ot,,,,is the standard


deviation. Thus, if an MT has a distance d from the BS, the
measured distance d,, is also a stationary random process with
mean d and variance (where c is the speed of light),
and the distance distribution is also normal [ I I ] , i.e.
du - Nl-d,(cotd,,l21 .
C. Vertical Handover Analytical Model
We notice that the measured criteria of signal level and
distance for both RAN cannot be directJy compared since the
monitored links come from different access networks, so
different thresholds of two access technologies are defined
respectively. Therefore, we assume R(/ and Rw are the RSS
handover thresholds on the satellite radio link and the WLAN
U
radio link, D(, and Dware the distance handover threshold.
Fig 3. Heterogeneous network integrating UMTS and WLAN Handover from the UMTS cell 1 to WLAN will occur when
the following relation is satisfied,

Because we give high priority to WLAN, handover in this


direction will occur without considering the RSS and the
distance criteria on UMTS cell link once condition (8) is
satisfied. Fig.4 illustrates the procedure of handover from
UMTS cell to WLAN.
On the other hand, when the active MT is using WLAN link,
handover from WLAN to UMTS cell 1 will occur when the
y r iand yw;are the correlation coefficients that determine the following relation is satisfied,
decaying rate of the correlation on different channels. do is the
distance between two locations used to calculate the correlation. [a< R,,] and

E. Distance Criterion

We know that MT's timing advance parameter [I21 is


HO criteria in this direction should consider the criteria of RSS
calculated based on the measurement of the signal delay
and the distance on UMTS link, in other words, the RSS on cell
between the MT and the BS. The delay measurement is
link should be beyond the handover threshold R(/ and the
generally composed of the real delay and measurement distance to BS should be below the handover threshold D(,
noise td,? .i We assume that measurement noise tdn is a when UMTS cell is to be the target network. Fig.5 illustrates
stationary zero-mean random process with normal the procedure of handover from WLAN to UMTS cell.

189

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:42:23 UTC from IEEE Xplore. Restrictions apply.
The 14'h IEEE 2003 International Symposium on Persona1,lndoor and Mobile Radio Communication Proceedings

. I C ~ IonI UMTS link I I Call on W L A N link 1

I-ianctover execution J

Fig.4 IJMTS-to-WLAN handover initiation procedure

D. Vertical Handover Probability


Based on (S), < l - - t M , ( ~ ~,) the probability to initiz 2 the
handover to WLAN on position x can be given by
ves Ir
I Handover execution I
Similarly based on (9), 4r;+ci(x),the probability to initiate Fig.5. WLAN-to-UMTS handover initiation procedure
the handover from WLAN to UMTS cell 1 on position x can
also be given by

4V4J(XI = 4 ( X I . p2 (XI (1 1)
2 ) When the user is far from the BS, i.e., DAp-Bs is
where f', (x)and p2(x) are calculated by
comparable with the distance threshold of UMTS link, the both
criteria of the UMTS cell should be considered because they
are both contributing factors for handover decision-making.
IV. NUMERICAL
RESUI TS
In Fig6 (a) and (b), we plot the probabilities of handover
from UMTS cell to WLAN based on (10). Handover
probabilities are calculated under two conditions, e.g.,
handover is either initiated only by RSS criterion, or initiated
Furthermore. in this handover direction when taking different by both RSS and distance criteria. The simulation parameters
D,-Ip-Bs into account we can obtain two scenarios as follows: from [IO] and [ 1 I ] are used. Fig. 6 (a) illustrates the handover
probabilities under the first condition against the distance
I ) When the distance D-Ip-Bs is far less than the distance between the user and AP as a variable. Fig.6 (b) illustrates the
handover threshold of UMTS link, that is handover probabilities under the second condition. By the
comparison of these two figures, we can observe that when
D, P- BS << Du (14) distance criterion is taken into account, the handover
the RSS and distance criteria for the UMTS cell are probabilities are smaller than those only based on RSS criterion.
As a resdlt, a trade-off between computational complexity and
unnecessary because the probability of [TJI(x) 2 R ( / ] is a smaller handover probability can be pursued by considering
the distance criterion or not.
almost always equal to 1 when the user is near to the BS. The
Based on (1 1-13) and (15), the probabilities of handover
same is true of the probability of [U',! 5 D o ] in this scenario. from WLAN to UMTS cell are calculated and the results are
So the handover condition in (1 1 ) reduces to a condition only shown in Fig7 (a) and (b). Probabilities are calculated
on the RSS on WLAN link. namely, according to different handover conditions: when DAP-BS is

190

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:42:23 UTC from IEEE Xplore. Restrictions apply.
The 14mIEEE 2003 International Symposium on Persona1,lndoor and Mobile Radio Communication Proceedings

09

021 l
o 200 400 GOO 800 iooo
Distance to WL4N A.P( x 0.1m )

(a) (b)
Fig.6. Probabilities of handover from UMTS cell to WLAN (a): RSS criterion; (b): RSS and distance criteria

0 35
x

5 03
=
0)

8 0 25
c
2 02
U
U
$ 0 15
A=
L
0 01
r

zh

0 05

Fig.7. Probabilities of handover from WLAN to UMTS cell (a): one RAN; (b): both RANs

far less than D[,, handover model only based on RSS on REFERENCES
WLAN link is used. Contrarily, handover model based on the
criteria of both RANs are used. Fig.7 (a) illustrates the ETSI, “Requirements and Architectures for intenvorking between
handover probabilities only based on the criterion of RSS on HIPERLAN/3 and 3rd Generation Cellular Systems,” Tech. Rep. EfSI
TRlOl 957, Aug, 2001.
WLAN link. Fig.7 (b) illustrates the handover probabilities
3GPP, “Feasibility Study on 3GPP System to WLAN Interworking”,
based on the criteria of two RANs. Comparing Fig.7 (a) with Tech. Rep. 3CPP TR 22.934 V I .2.0, May 2002.
Fig.7 (b), we can draw the following conclusions: when K. Pahlavan, P. Krishnamurthy and A. Hatami,” Handoff in IHybrid
D4p-Bs is far less than threshold Dli, the probability Mobile Data Networks”, IEEE Personal Comnz. Mag., April 2000.

I-
ot ti,, ( A) 5
1
R,,, dominates the handover probability fh,+(,
Fig 7 (a). When D I l p T B ~is comparable with D(,, that is, the
(x) in
K. Salkintzis, C. Fors and RS Pazhyannur, “WLAN-GPRS Integration
for Next Generation Mobile , Data Networks,” IEEE Wireless
Connnunicaiions. pp I 12- 123, October 2002.
Jean Tourrilhes and Casey Carter. “P-Handoff: A protocol for fine
user is far from the BS of UMTS cell, the cell link may provide grained peer-to-peer vertical handoft“, Hf Labs Technical R e p m s
low signal quality for the user. So the handover to UMTS cell www.hpl.hp.com/techreports/2002/, HPL-2002-6 I , 2002.
may be discouraged by incorporating the criteria of the cell link, M. laseemuddin, An Architecture. for Integrating UMTS and 802.1 1
which I S shown with smaller probabilities in Fig.7 (b) WLAN Networks , IEEE Symposium on Computers and
Conzrnunrcations (ISCC 2003), July, 2003
compared with those in Fig7 (a).
3GPP, “UTRAN Overall Description”, TS 25.401 V3.3.0, June,2000.
v. CONCLUSION MStemm and RH Katz, “Vertical handoffs in wireless overlay
networks,” Mobile Nehvorks and Applicairons 3(4) 1998 pp.335-350.
An integrated architecture of UMTS and WLAN has been N. Tripathi, “Generic Adaptive Handoff Algorithms using Fuzzy Logic
described. in which a network intenvork unit (NIU) is and Neural Networks”, Ph.D. Thesis, Virginia Polytechnic Institute and
introduced to hide the WLAN particularities. Based on the State University, August 1997.
advanced analytical vertical handover model, the paper focuses Amir Majlesi, Babak H. Khalaj, “An Adaptive Fuzzy Logic Based
011 the performance evaluation of the handover probability. Handoff Algorithm For Interworking Between Wlans And Mobile
Networks”, IEEE PlMRC 2002
Numerical results show that our model can match to the real
case well. Furthermore, relationships among handover Wei Zhao, R. Tafazolli, B.G. Evans. “Internetwork Handover
Performance Analysis in a GSM-Satellite Integrated Mobile
performances criteria will be established to get a Communication system”, IEEE journal on selected areas in
comprehensive evaluation, such as handover probability, communications, vol. 15, no. 8, pp 1657- I67 1,October 1997
uiinecessaiy handover rate, average number of handovers, and ETSI-European T e l e c o m m u n . Standardization Inst., GSM
call-dropping probability, etc., this is our work in progress. Recommendation 05.08.

191

Authorized licensed use limited to: National Ilan Univ.. Downloaded on March 04,2021 at 15:42:23 UTC from IEEE Xplore. Restrictions apply.

You might also like