Sunteți pe pagina 1din 5

Design of Optimum Parameters for Handover Initiation in WCDMA

Jongin Kim", Dong-hoi Kim**, Pyeong-jung Song**, Sehun Kim*


*Korea Advanced Institute of Science and Technology (KAIST)
373-1, Kusung dong, Yusung Gu, Taejon, 305-701, KOREA
**Electronics and Telecommunications Research Institute (ETRI)
161 Kajong-Dong, Yusong-Gu, Taejon, 305-350, KOREA
Email :jikim@tmlab.kaist.ac.kr
Abstract- WCDMA handover algorithms employ signal
averaging, hysteresis and the Time-to-Trigger mechanism to
optimize the trade off between number of unnecessary handover,
reported events(system load) and handover delay time. We
investigate optimal parameters for the WCDMA intra-.
frequency handover algorithm and the impact of each
parameter on the system performance. Number of reporting
events triggered for handover and handover delay are key
performance measures in this paper. The study shows various
tradeoffs between the parameters related to averaging,
hysteresis and Time-to-Trigger. We have also discovered that
the layer3 filter and Time-to-Trigger mechanism may cause
negative effects on each other in some cases and there are
optimum values, when used simultaneously.

1. INTRODUCTION
Design of handover initiation can be made to use several
measurements such as the received signal level from the
communicating and neighboring base stations, the path loss
to the base stations, and bit error rate. In general, hysteresis
and signal averaging is employed to enhance the performance
of handover(i.e. probability of unnecessary handover at the
expense of handover delay). Previous studies on handover
initiation have revered that there are trade offs between
handover delay and number of unnecessary handover.
Handover initiation criteria analyzed in literature are
mainly based on the length of averaging window, the
threshold level and the hysteresis margin. In addition,
WCDMA introduce the Time-to-Trigger mechanism to
reduce unnecessary signaling and ping pong effects. Also,
averaging window is used to smooth out random signal
fluctuations and to make handover decisions to be based on
underlying trends and not instantaneous changes.
Soft handover is essential for intra frequency in WCDMA.
The active set i s defined as the set of base stations to which
the mobile users is simultaneously connected. Soft hand over
involves active set update procedure which include signaling
of appropriate event triggered by the mobile based on the
measurement of the measurement quantity (i.e. Ec/Io, path
loss, etc). Frequent reporting may cause unnecessary
handover and signaling overload. On the other, if the
reporting is too seldom, it may increase the handover delay.
WCDMA(3GPP) recommendation does not specify the
measurement and averaging interval be fixed or variable.
Actual physical layer measurement depends on the
implementation of the mobile unit. However, WCDMA
specifies the network controlled features to enhance the

0-7803-7005-8/01/$10.00 0 2001 IEEE

performance, which include the hysteresis, Time-to-Trigger


and Layer3 filtering. A network controlled Layer3
fiItering(exponentia1 smoothing) provides same options as to
hysteresis and Time-to-Trigger to some extent, but give some
extra benefits which makes it possible to control the rate of
reporting, i.e. system loads. Therefore, it is our interest is to
investigate the impacts of each of the network controlled
elements, including Layer3 filter, hysteresis margin and
Time-to-Trigger, to handover performance after applying a
minimal physical layer measurement. Our goal is to optimize
the parameters for these handover mechanisms considering
various ttadeoff relations. By using an appropriate
combination of filter, hysteresis and Time-to-Trigger, it is
possible to fine tune the real time decisions to be optimal in
time and amplitude. Therefore, we can optimize parameters
related to hiandover decision.

3. SYSTEM DESCRIPTION

A. Measurements and Signaling


In WCDMA system, the mobile station performs intrafrequency measurement and sends measurement report to the
Radio Network Controller(RNC), where the final decision is
made about which cell to add or remove from the Active
Sets[2]. The intra frequency measurement is done on the
downlink P-CPICH[l]. Measurement quantity can be any of'
the followings; Ec/Io, path loss and the Received Signal Code
Power[ 11.
Consider a model for a network controlled handover
filtering (Signal averaging) shown in Figure 1. This model is
as recommended in 3CPP specification[ 11. Parameter 1 is
related to :shape of Layer3 filter provided by the network and
parameter 2 is related to types of handover, i.e. intra
frequency, inter-frequency, etc and reporting criteria.

2768

Layer I

filtering

Pamters 1

Paramters 2

Layer 3

% filtering -b Evaluation

_
.

ofreporting

Figure 1. Model for handover measurements

D
*

Uu

lub

LT'

Lr'
UE-RRC

Physical layer implementation (inputs A and Layer 1


filtering) is not constrained by the standard i.e. the model
does not state a specific sampling rate or even if the sampling
is periodic or not. What the standard specifies is the
performance objectives and reporting rate at point B in the
model. The reporting rate is equal to the measurement period,
which is 200ms for intra-frequency measurement. The
performance objectives for the physical layer measurements
are specified in [3].

UE-L1

MEASUF

E5
JNTROL

I
CPHY-Measurement-REQ

a
Measurement

CPHY-Measuremenl-IND

In addition, the Layer3 filtering is performed according to


the following exponential averaging formula to give more
accuracy.

Measurement

CPHY-Measurement-IND

cntena
fulfilled

I
MEASUREMENT REPORT

The variables in the formula are defined as follows; F,, is the


updated filtered measurement result. F,+, is the old filtered
measurement result. M,, is the latest received measurement
result from physical layer measurements. If a is set to 1 that
will mean no layer 3 filtering. Also, smaller a will mean that
it is giving more weights to past samples.
Hysteresis and Time-to-Trigger mechanism on the other
hand is important for reducing unnecessary signaling or
handover and they complement to averaging mechanism.
Evaluation of reporting criteria is based on the measurement
results (after appropriate filtering) using the hysteresis and
Time-to-Trigger mechanism. The reporting event 1 A and 1 B
is defined as;
-Meas_Sign>Best-Ss-Hyst-Add
-Meas-Sign<Best-Ss-Hyst-Drop

Figure 2. Reporting of Measurement

AT

for AT: Event I A


for AT: Event 1 B

is
If Measured quantity at point C(Meas-Sign)
continuously larger than the best measured set present in the
active set( Best-Ss) minus add hysteresis for Time-toTrigger(AT), then measurement reporting message for the
eventlA is sent over the air interface to the RNC as shown in
Figure 2. Similarly, the eventlB can be reported. The
reporting events constitute basic input to handover algorithm
in the RNC, where the handover decision is made (i.e. Active
set addition, Active set removal).

Figure 3. Simple handover algorithm

4. SIMULATION MODEL

B. Handover Scheme
An example of basic handover algorithm is shown in
Figure 3, which exploits the hysteresis and Time-to-Trigger
mechanism. Event 1A and event 1 B are implemented in this
example. Since the soft handover algorithm is performed at
RNC, load control strategy and other radio resource
management strategy can be exploited for active set updates
considering any of the measurement quantities described in

PI.

We consider two cells each with radius of 2000m and


mobile is allowed to move from BSI to BS2 in a straight line
joining them (Figure 4). Our measurement model is identical
to that shown in Figure 1. We apply a basic Layerl filter,
which takes CPICH RSCP samples every lOms at input A
and then 20 samples are averaged over a rectangular block for
the duration of measurement period (i.e. 200ms). A typical
signal output of Layerl filtering in our model, which in
practice depends on the implementation of mobile handset, is
shown in Figure 4.

2769

2:

r ( d ) = K, - K , log,,(d) + v ( d ) + 201og,,[e(d)]

The parameters K1 and K2 accounts for path loss, v(d) i s


the shadow fading process; zero mean, variance lOdB,
stationary Gaussian process. The shadowing process is
assumed to have the exponential correlation function
proposed by Gudmundson[6]. Decorrelation distance is
assumed tci be 20m in vehicular environments [4].

I
0

1000

2000

3000

4000

For the fast fading, we use ITU Vechicular A model[4].


Received s,ignal after filtering is then given by

Distance traveled by mobile ( m )

Figure 4. Model for handover measurements

F ( d ) := K, - K, l o g , , ( d )

As it can be seen, the basic Layerl filtering of 200ms in


our model does not completely average out the signal
fluctuation. Relationship between the accuracy and the
measurement distance is described in [5]. Our interest i s to
investigate the effects of the network controlled elements,
such as Layer3 filter, hysteresis margin and Time-to-Trigger,
on the handover performance after applying a minimal
Layerl filtering. The simulation parameters are listed below
and the channel model i s described in the following
subsection.

Parameter

Value

Propagation Model

128.1+37.6log (R)

Channel Model

ITU-Vehicular A

Standard Deviation of
Log-Normal fading
Decorrelation destance

I
I

B. Performance Measures
Optimal handover is the trade off between the number of
unnecessai-y handover and the handover delay. Many
previous literatures have studied the properties of this trade
off for various parameters such as the hysteresis margin and
the length of averaging distance[8]
In WCDMA, the standards specify the measurement model
and the range of parameters like Layer3 filter coefficient,
hysteresis and Time-to-Trigger. But, the impacts of these
parameter:: and different choices for the values remain to be
clarified.
WCDMA use soft handover mechanism to enhance the
coverage and capacity of the network. Soft handover
mechanisrn involves active set update and removal as
described in the previous section. To many reporting events
will cause unnecessary active set updates and increase the
signaling load. On the other hand, infrequent reporting may
cause delay in handover. Optimal size of soft handover
depends on loading conditions and, etc. The size of soft
handover area can be also controlled by the system
parameters.
Number of reporting events triggered for handover and
handover delay are key performance measures in this paper.
In our simulation, the tradeoff between the number of
reporting events and average distance of active set
additioniremaval, averaged over 1000 runs, are investigated
with different hysteresis margins, Layer 3 filter coefficients
and Time-to-Trigger. Average distance of reporting event 1A
i s the mean distance at which the active set addition for BS2
takes place.

1OdB
20m
47dBm

CPICH Power

I
I
I

H yst-ADD

lSdB,3dB,4SdB,6dB

Hyst-Drop

2SdB,5dB,7SdB,lOdB

Ti me-to-Tri gger

Oms,200ms,400ms

Measurement Pen od

200ms

Sampling interval
Layer3 filter coefficient

-~

+ V ( d )+ 2010gl,[e(d)]

0.1-1

Table 1. Simulation parameters


A. Propagation Model
The received signal at a mobile consists of three parts; path
loss, slow fading and fast fading (Rayleigh distributed).
Therefore, the received signal (in dB) as a function of mobile
distance i s given by,

5. SIMULATION RESULTS
Figure 5 . shows the expected number of reporting eventlA
for mobile traveling at speeds 50km/h and 120km/h with
various hysteresis, not using Time-to-Trigger. Number of
reporting events i s quite large when Time-to Trigger is not

2770

used. It can be observed that the layer3 filter can reduce the
number of reporting events significantly. Especially at low
mobile speed, it shows significant improvements. The effect
of the hysteresis is also shown in this figure. Figures 9. and
IO. show the mean distance at which the mobile sends the
reporting EventIA for BS2 and eventlB for BS1,respectively.
It can be interpreted as the expected point where the mobile is
enterindleaving the soft handover area. This position
depends on the setting of hysteresis levels, but Layer3
filtering also has effects of delaying the distance of
enterindleaving the area. Similarly, as show in Figures 11
and 12, the Time-to-Trigger mechanism also delays the
reporting events. The gain of soft handover and optimum size
depends on many factors including the system loads and the
capacity.
Figure 6. represents the expected number of reporting
event 1A with Time-to-Trigger of 200ms. It is interesting to
observe that the rate of increase of the number of reporting
event starts to slow down at some point as alpha is increased.
Further, in Figures 7 and 8, with 400ms Time-to-Trigger, the
number of reporting actually begin to decrease at some point
along alpha. This characteristic is explained as follows. First,
with no Time to Trigger, the smoother curve will obviously
give less reporting events since it has smaller variations. If
the Time-to-Trigger of 200ms and 400ms is used, two and
three adjacent samples are subsequently evaluated,
respectively. Exponential averaging induces correlations
between these samples. Correlated samples will be
undesirable than independent samples in extracting the
average value. Therefore, there exists a tradeoff between
obtaining the stable measurement results and getting
independent samples. Consequently, in setting Time-to-

Trigger and Layer3 filter constant, we may consider this to be


compromised.
60

I
4 Hyst-Add=B, Hyst-D rop=lO,

50

-0

4 1

00

02

/J

06

04

10

08

Layer3 Filter Coefficient(a)

Figure 6. Average number of reporting event 1A


(Time-to-Trigger = 200ms )
15

.+

1 3 -

0,

c
-

r
2 1 2 a,
[r

1 1 -

-e Hyst_Pdd=3, Hyst_Drop=5
0 Hysl-Pdd=l 5, Hyst_Orop=25

10-

09

00

02

04

06

08

10

12

Layer3 Filter Coefficient(0)

Figure 7. Average number of reporting event 1A


(V=120km/h, Time-to-Trigger = 400ms)

V=l20 krnh

. H yst-Pd d=3, Hyst-D rop=5, V=l2Okrn/h


Hyst-0 ro p=lO , V=SOkrn/h
Hyst_Nd=3. Hyst_Drop=5, V%Okrn/h

t H yst-Ad d=6,

"5

40

f---

ti5

'U

'v

0)

30
0.

20

0
L

a,

5
Z

10
0'

.- 0

I
00

02

04

06

08

10

Hyst_Pdd=3, Hysl_Drop=5

t Hyst-Add=l 5. Hyst_Drop=25

I
12
00

Layer3 Filter Coefficient(a)

02

04

06

08

10

Layer3 Filter Coefftcient(0)

Figure 5. Average number of reporting eventlA


(Time-to-Trigger = Oms)

Figure 8. Average number of reporting event 1A


(V=50km/h, Time-to-Trigger = 400ms)

277 1

12

1600

I \

1750
1700

2300

.-.
E

o\

2280

\
0

Time_to_Trigger=2SOms

2269

1650

-c H yst-Ad d 4 d B , Hy st-D rop=5dB


1600

Hyst_Pdd=6dB, Hyat-Drap=lOdB

2240
A
.

Ln

1550

0.

22zD

2200

c
00

02

04

06

OB

10

12

Layer3 filter coefficient(a)

Figure 12. Mean distance of reporting eventl B(50km/h)


(Hyst-Add=3, Hyst_Drop=5, 50km/h)
2600
0

-E

--

2500

2400

.E.

2300

Hyst_Pdd=3, Hyst_Drop=5
Hyst_Pdd=6, Hyst-Drop=lO

i
5

2200

21 00

02

00

04

06

08

10

12

6. CONCLUSIONS
This study investigates the impacts of each of the network
controlled features (including Layer3 filter, hysteresis margin
and Time to Trigger) in handover initiation mechanism. The
study shows various tradeoffs between the parameters. It was
investigated in terms of the number of event reporting and
handover delay distance. The performances are also shown to
depend on the velocity of the mobile. We have also
discovered that the layer3 filter and Time-to-Trigger
mechanism may cause negative effects on each other in some
cases anti there is an optimum combination, when used
simultaneously. The results presented in this study may help
in understanding the behavior of the features related to
triggering of handover measurement reports and in extracting
optimum parameter values. Further, these results can be used
for developing more efficient handover algorithms.

Layer3 filter soefficient(0)

Figure I O . Mean distance of reporting e v e n t l B(50km/h)


1820

-E
-

1800

-4- Time-to-Trigger=Oms
0 Time-to-Trigge~200ms
t Time-to-Trigge~400ms

1780

...

1760
\O

1740

1720

00

02

04

06

08

10

12

Layer3 filter coefficient(a)

Figure 11. Mean distance of reporting eventlA


(Hyst_Add=3,Hyst_Drop=5, 50krn/h)

2112

REFERENCE

[ I ] 3GPP TS 25.302 Ver 3.3.0 Services prvided by physical


layer, March, 2000
[2] 3GPP TS 25.331 Ver 3.6.0 RRC protocol specification,
March, 2000.
[3] 3GPP TS 25.133 Ver 3.5.0 Requirements for Support of
Radi o Resource Management ,December, 2000
[4] ETSI TR 101 112 V3.2.0, Selection procedures for the
choice of radio transmission technologies of the UMTS,
April, 1998
[ 5 ] C.Y.Lee, Mobile Communications Engineering
McGrawHill, 1981
[6] M.Gudmunson, Correlation Model for Shadow Fading in
Mobile Radio Systems, Electronics Letter, Vol 27, no23,
pp 2145-2146, NOV 1991.
[7] R.Vijayan and J.M.Holtzman, A Model for Analyzing
Handoff Algorithms, IEEE Trans. On Vehicular
Technology, August 1993.
[8] Gregory P. Pollini, Trends in Handover Design, IEEE
Cominunications Magazine, March 1996

S-ar putea să vă placă și