Anda di halaman 1dari 8

978-1-4244-1870-1/08/$25.

00 2008 IEEE

AbstractIn femtocell deployments, leakage of the pilot signal
to the outside of a house can result in a highly increased signalling
load to the core network as a result of the higher number of
mobility events caused by passing users. This effect can be
minimized by reducing the pilot power. However this approach
could lead to insufficient indoor coverage, which also causes an
increase in mobility events. In this paper, coverage adaptation is
proposed for femtocell deployments that uses information on
mobility events of passing and indoor users to optimize the
femtocell coverage in order to minimize the increase in core
network mobility signalling. Different coverage adaptation
methods are discussed and it is shown that the proposed mobility
event based self-optimization can significantly outperform
simpler methods that aim to achieve a constant cell radius.
I. INTRODUCTION
A significant interest within the telecommunications
industry has recently focused on the subject of femtocells.
Their commercial potential has been analysed and a recent
forecast by ABI Research predicts that the market for
femtocell equipment will grow 95 percent per year, to more
than $4 billion by 2012 [1]. The interest in this technology
area occurs for many reasons, as a means of extending
coverage into residential areas, to focus on technologies that
can potentially be used for targeting particular market
segments more effectively, or as a way of affecting a fixed
mobile convergence strategy. This list is by no means
exhaustive. However, it would be fair to say that three
technical factors are key to the success of femtocell
technology: the guaranteed coverage area of the cell, the auto-
configuration and self-optimization capabilities of the cell, and
the core network signalling caused by user mobility due to
capture effects.
It should be noted that femtocell technology is still in
development and key aspects of the necessary technology have
been studied at length and reported in the literature. For
example, since the deployment of femtocells can be in orders
of magnitude more numerous than traditional cellular
deployments, architectural aspects had to be re-evaluated and
more IP-like architectures proposed [2] along with mobility
enhancements [3] and paging mechanisms for distributed
environments [4]. In parallel with this, the autonomic nature
of such deployments and their evolution have also been
studied. See for example the initial studies on self organisation
of wireless networks [5][6][7][8] and the development of these
ideas to more implementable means for co-channel
deployment [9][10][11][12]. The context of self-deployment
has even been extended to include future scenarios which, as
this technology matures, will be the next natural step in terms
of the application of this technology to cognitive approaches
[13]. The further extension of complete autonomic methods to
wider scale deployments (not just femtocells) has been
proposed in [14]. Furthermore, studies on the financial impact
on the application of this kind of technology on current
cellular systems have also been undertaken [15].
While the above indicates that many of the major issues
have been studied, there are still many areas that are left open
and as the technology matures are revisited, especially where
co-channel deployments are concerned. One such area is the
capture effects that femtocells will have on mobile terminals
that are moving through a macro coverage area, both in active
and idle mode (i.e. the unnecessary mobility events triggered
by transient mobile terminals moving to a femtocell) [12]. To
some extent these are being addressed in the LTE/SAE
evolution of the current 3GPP standard via the creation of
equivalent paging areas [16]. However, pragmatically, until
revisions of the current 3GPP standard occur or newer more
targeted standards emerge that deal specifically with the nature
of femtocell deployments, there will have to be concerted
efforts on the issues of capture effects in practical femtocell
deployments, since femtocells will be in common deployment
for some years before these features are adopted. Therefore it
is necessary to seek solutions to mitigate capture effects in a
general sense. In this paper a novel approach for coverage
optimization is proposed that favourably reduces this effect.
The paper is organized as follows. Section II provides an
overview of femtocell technology. In Section III, auto-
configuration and self-optimisation strategies are presented
that are directly applicable to a femtocell environment. In
Section IV the relative performance of these strategies is
reported and discussed, while Section V concludes with some
final remarks.
II. FEMTOCELL OVERVIEW
Femtocells are defined broadly in this context as low-cost,
low-power cellular base stations that operate in licensed
spectrum to connect conventional, unmodified mobile
terminals to a mobile operators network. The range of
femtocells are in the tens of meters. They utilise broadband
DSL or cable Internet connections for backhaul back to the
operators core network. Femtocells were initially designed for
residential use, wherein improved and personalised indoor
cellular coverage is provided. However, the self-optimisation
Self-optimization of Coverage for Femtocell
Deployments
Holger Claussen, Lester T. W. Ho, Louis G. Samuel
Bell Laboratories, Alcatel-Lucent, Swindon, UK.
{claussen, lho1, lsamuel}@alcatel-lucent.com




and coverage principles can be extended beyond this initial
deployment scenario to include campus, enterprise and
metropolitan zone deployments. Mobility and handovers
between the macrocell underlay and the femtocell overlay are
implemented seamlessly from the end-users point.
Fig. 1 shows an overview of a femtocell architecture.
A. Femtocell value proposition
From a network operators standpoint, femtocell
deployments offer many advantages. One benefit is it allows
operators to offload significant amounts of traffic away from
the macrocell network, as a large portion of calls made by
users on cellular networks takes place indoors at home or
while at work. In certain circumstances, it has been shown that
the traffic offload to femtocell deployments with slightly
increased coverage can lower the costs of network operators
by up to 70% [15]. This offloading is especially beneficial
when high data rate services are considered.
Femtocells provide an opportunity for the provision of
ubiquitous high-speed data access. It has been widely argued
that the shift towards the use of small cells is a route to
providing the near infinite bandwidth required in the future
[17]. Small cells are well suited for the provision of high data
rate services as it offers the localisation of radio signals, such
that the shared use of bandwidth in macrocells can be replaced
with a more personal use of bandwidth. The implication of
the personalisation of mobile cellular access is the potential to
create new consumer offerings that take advantage of the long
tail distribution of consumer choice [18] thereby increasing
consumer usage and reduce churn.
The fact that femtocells works with unmodified mobile
terminals is also an important advantage over other competing
technologies such as UMA, that requires terminals capable of
working with multiple access technologies. Such terminals are
more expensive and the power requirements for maintaining
two or more radio interfaces can reduce its battery life.
Small cells can also use higher frequency bands that are
more suited to provide high data rates since the geometry of
deployment can assist in the mutual isolation of the femtocells.
So while the current use of femtocells is limited to residential
use, it can be seen as an important step in the roadmap towards
the provision of high data rate mobile services.
B. The requirement for auto-configuration and self-
optimization in femtocells
The number of femtocells involved in a wide-scale
deployment will be orders of magnitude more than the
numbers commonly associated with cells used in macro
cellular deployments. With such large numbers, the
conventional manual approach to cell-planning and
deployment of base stations would not be feasible due to the
resulting prohibitively high costs. Femtocells therefore have to
be deployed by the end-users themselves, and the femtocells
must be able to auto-configure all the required parameters and
self-optimize them during operation with minimal human
intervention.




Fig. 1. Overview of femtocell architecture

In this paper, techniques for the auto-configuration and self-
optimization of femtocell coverage are presented. Being user-
deployed, the operator no longer has control over, amongst
other things, the location in which femtocells are deployed.
This can be problematic because a femtocell can be deployed
in an unsuitable location, for example next to a window facing
a sidewalk which causes a lot of its radio signals to leak
outside the home. This causes unwanted interactions with
public users outside, which can result in excessive handover
and mobility signalling [12]. As a result, it is important that the
femtocell is able to automatically adjust its coverage area such
that it does not leak outside into public areas, but at the same
time provide adequate coverage in the home.
III. AUTO-CONFIGURATION AND SELF-OPTIMIZATION
STRATEGIES
This paper distinguishes between auto-configuration and
self-optimization schemes. The difference is that auto-
configuration schemes provide an initial configuration whereas
self-optimization schemes optimize the current configuration
during operation, for example based on measurements or
statistics collected over time.
A. Auto-configuration schemes
The following pilot power auto-configuration schemes are
considered in this paper:

(1) Fixed power. While this cannot be considered as a real
auto-configuration scheme, using a pre-configured, fixed
power is investigated for comparison. For the discussions in
this paper, a fixed pilot power of -10dBm is considered. This
configuration achieves in 90% of all possible femtocell
locations on average a target range of 10m, which results in



full indoor coverage for the considered house model when the
femtocell is deployed in the centre of the house.
This approach has the disadvantage that it results in a highly
variable femtocell radius for co-channel deployments ranging
from less than the target cell radius when deployed close to the
macrocell up to several times the target radius when deployed
at the macrocell edge. Therefore, it is not suitable for co-
channel femtocell deployments.

(2) Distance based. The femtocell pilot power is configured
such that it is received on average with equal strength as the
pilot power received from the strongest macrocell at a defined
target cell radius of r, subject to its maximum power P
pilot,max

[11]. The initial femtocell pilot power can be calculated in
decibels as

P
femto,pilot
= min(P
macro,pilot
+ G
macro
L
macro
+ L
femto
(r), P
pilot,max
), (1)

where L
femto
(r) is the estimated path-loss from the femtocell to
a UE at the target femtocell radius r. L
macro
denotes the path-
loss between the macrocell and the femtocell. P
macro,pilot
is the
pilot power transmitted by the macrocell and G
macro
is the
macrocell antenna gain in direction of the femtocell. The path
loss between a transmitter and a receiver separated by a
distance d can be modelled as

L(d) = L
1
+ L
2
10 log
10
(d). (2)

For free-space loss L
1
and L
2
are 38.5dB and 2dB respectively,
for typical urban path loss L
1
and L
2
are assumed to be 28dB
and 3.5dB respectively. Optionally, additional estimated wall
losses can be considered.
This achieves a roughly constant cell range independent of
the distance to the macrocell in the co-channel hierarchical cell
structure considered in this paper. Note that variations are
caused by the varying shadow fading losses that cannot be
taken into account by the path-loss model.
While this approach is suitable for the initial power
configuration for femtocell deployments it has the
disadvantage that a large amount of information on the
macrocellular network, such as cell locations, power levels,
antenna orientation and gain, is required. The approach also
relies on a reliable path loss model for L
macro
.

(3) Measurement based. The femtocell pilot power is
configured using the same principle as above with the
difference that the received power P
Rx-pilot,macro
is not estimated
using a path-loss model, but measured using the built in
measurement capability of the femtocell. Then the initial
femtocell pilot power can be calculated in decibels as

P
femto,pilot
= min(P
Rx-pilot,macro
+ L
femto
(r), P
pilot,max
). (3)

The path-loss L
femto
(r) from the femtocell to a UE at the
target femtocell radius r can be modelled as described above
in (2).
As with the distance based auto-configuration, this approach
achieves a roughly constant cell range independent of the
distance to the macrocell in a co-channel hierarchical cell
structure. Note that variations are caused by the varying
shadow fading losses that have an impact on P
Rx-pilot,macro
, but
are unknown to the femtocell.
This approach has the advantage that no information on the
macrocell network is required. However, since the received
signal from the macrocell P
Rx-pilot,macro
depends highly on the
deployment location and the resulting indoor wall losses, this
causes some variability in the achieved cell range.
B. Self-optimization
During operation, information on the number of mobility
events and their time are collected and used for further
refinement of the coverage by self-optimization of the pilot
power. The pilot power can be initialized using any of the
auto-configuration schemes described above. Here, the
measurement based auto-configuration process is used for
initialization due to its simplicity and the ability to achieve a
target range independent of the distance to the co-channel
macrocell. The proposed self-optimization methods are
described below.

(1) Mobility event based adapted to minimize mobility
events of passing users. This optimization method takes only
unwanted mobility events from transient users into account
that briefly hand over to the femtocell and hand back
immediately after passing the house. The objective is to
maximize the indoor coverage under the constraint of limiting
the mobility events from such passing users. Starting from a
measurement based pilot auto-configuration, the femtocell
counts mobility events and classifies them into wanted and
unwanted events over time, dependent on whether the mobile
is registered at the femtocell and on the time a UE spent in the
cell before moving back to the macrocell.
If the number of unwanted mobility events of passing users
exceeds a pre-defined value of n
1
events per time t
1
the
femtocell reduces its pilot power by a step
1
. The femtocell
then starts a new mobility event count for the updated
configuration.
If the number of unwanted events is smaller or equal to a
pre-defined acceptable value of n
2
events for a time t
2
, the
femtocell increases its pilot power by a step
2
, to provide
improved indoor coverage and starts a new mobility event
count for the updated configuration.
The measurement intervals and the optimization process is
illustrated in Fig. 2.
In this paper it is assumed that all mobility events of passing
users shall be prevented, therefore n
1
= n
2
= 0. The
measurement interval t
1
is equal to the optimization iteration
time, and for any occurring event the femtocell pilot power is
reduced by
1
= 3dB. In order to speed up the cell radius
increase after the initial auto-configuration, t
2
is set to a low
value of 120s until a first unwanted mobility event is detected.
Then t
2
is increased significantly to 6h so that the coverage is
dominated by the decreases caused by unwanted mobility
events. The step size
2
for the pilot power increase was
selected to be 0.3dB to alter the cell size only in small steps.
While the selected parameters perform well in the considered
scenario, automatic optimization of these parameters is an item
for future studies. Such optimized parameters would have to



result in a fast convergence to the point with the best trade-off
between femtocell coverage and increase in core network
signalling for the operator.

(2) Mobility event based adapted to minimize the total
number of mobility events. Instead of utilizing only unwanted
mobility events from transient users, this method tries to
minimize the total number of mobility events. Initialized using
the measurement based pilot auto-configuration, the self-
optimization process is started as described above taking only
mobility events from transient users into account until
converged. The convergence point is detected by a change of
direction from decreasing the pilot power directly followed by
an increase.
From this point on all mobility events are used for further
optimisation and the pilot is increased step by step as long as
the resulting total number of mobility events per time reduces
until a minimum is reached. While this technique does not
necessarily find the global minimum as the optimum method
described below, it can sometimes further reduce the total
number of mobility events compared to the method that takes
only passing users into account.
To prevent getting stuck in a local minimum, instead of
increasing the power step by step, higher increases can also be
tried after the minimum has been reached to check if there is
an alternative lower minimum achievable with higher power.
However, since this delays the optimization process and
potentially results in an increased number of mobility events
during this process, it is not considered here.

(3) Mobility event based optimum (Benchmark). The
optimum pilot power that achieves a coverage that on average
minimizes the total number of mobility events can be found by
an exhaustive search over all possible pilot power settings
starting from the lowest value that results in no outdoor
mobility events to the maximum allowed pilot power value
(here performed with step size of 1dBm) given the assumed
user mobility models. Starting from the lowest value that
results in no outdoor mobility events guarantees some indoor
coverage and prevents the method from disabling the pilot
completely (this would always result in no mobility events).
This method is impractical in reality due to the long
optimization time of the exhaustive search and the fact the
many poor values are chosen during the search, which results
in an increased amount of mobility signalling during the
optimization. However, this method is valuable for comparison
to show how far other more practical methods perform
compared to the optimum.

Examples and discussions. Fig. 3 and Fig. 4 show examples
for the number of resulting mobility events from indoor and
outdoor users for two different femtocell locations within the
house.
Fig. 3 shows examples for a typical deployment where the
femtocell is located in the centre of the house. It can be seen
that in this case an optimization with method 1 that when
converged selects the maximum pilot power that does not
cause outdoor mobility events (2dBm) is not optimum in terms


Fig. 2. Overview measurement intervals and mobility event
based coverage optimization that minimizes mobility events of
passing users


Fig. 3. Example of mobility events with different pilot
power values for a femtocell deployed in the centre of a house
and infrequent outdoor traffic


Fig. 4. Example of mobility events with different pilot power
values for a femtocell deployed near the front window and
infrequent outdoor traffic
collect measured mobility
events occurred during
measurement intervals
t1 and t2
detect unwanted mobility events
nt1, nt2
Decrease coverage:
Pfemto,pilot = Pfemto,pilot 1,

reset mobility measurements,
set t2 to high value.
Increase coverage:
Pfemto,pilot = Pfemto,pilot + 2,

reset mobility measurements.
nt1 > n1
nt2 n2
continuous optimization iterations
time
t1
t2
mobility events
yes
yes
no
no
now




of minimizing the total number of mobility events, but still
achieves a good result. This example also shows a case where
method 2 would be stuck in a local minimum at a power of
2dBm when only increases of 1dB are tried. If a 2dB increase
is tried it achieves the global optimum. Method 3 finds the
global optimum at a pilot power of 4dBm.
Fig. 4 shows a special case where the femtocell is deployed
in an unsuitable location close to a window near a footpath.
Here method 1 would converge to a pilot power of -23dBm.
Method 2 is stuck at the same value as method 1. Method 3
finds the optimum value at 9dBm and provides full coverage
for the indoor user. However, from a global network
perspective it might still be better to optimize according to
methods 1 or 2 since this would encourage the femtocell owner
to re-deploy the femtocell to a better location inside of the
house which would give him better indoor coverage and result
in low numbers of mobility events for passing users.
While the proposed mobility event based self-optimization
methods above are investigated for a co-channel scenario with
public access, they can also be applied to other scenarios such
as for example operation on separate channels with private
access. Here, instead of unwanted mobility events (i.e.
handovers), rejected handover or camping attempts can be
used. Note that the self-optimization methods described above
are not specific to the assumed mobility model in this paper.
In a real network, typically large variations of traffic are
encountered dependent on the time of the day. This is not
considered in this paper, but can be taken into account by
optimizing power levels independently for different times of
the day using any of the self-optimization methods above.
IV. PERFORMANCE ANALYSIS
The performance of the proposed self-optimization methods
is analysed using system level simulations. This section
describes the user modelling, the simulation assumptions and
the results.
A. User modelling
To simulate the behaviour of the femtocell user, an indoor
user mobility model is used. The indoor model uses a
collection of waypoints defined throughout the house model of
a terraced house, shown in Fig. 5. The femtocell user moves
from different rooms within the house and spends a certain
amount of time at each waypoint in the room before moving on
to another room at a speed of 1ms
-1
. Each room has a
probability that it will be visited by the user, with the lounge



Fig. 5. Waypoints of the indoor user mobility model

Fig. 6. Distribution of time spent in each room


(a) (b)

Fig. 7. Coverage and received pilot power levels [dBm]
around one femtocell for (a) macrocell-, and (b) femtocell
access for r = 10m (free-space)

TABLE I: SIMULATION PARAMETERS
Outdoor path loss is modelled as 28+35log
10
(d) dB where d is the
distance from the base station in metres.
Indoor path loss is modelled as 38.5+20log
10
(d) + L
walls
dB where
the wall loss L
walls
is explicitly modelled (15/10/7 dB for
external/internal/light internal walls respectively, 3 dB for doors,
and 1 dB for windows)
Shadow fading is modelled as random process with log-normal
distribution (8 dB standard deviation for the macrocell signal
where other houses and obstacles are implicitly modelled, and 4
dB standard deviation for the femtocell signal)
The receiver noise power is modelled as 10 log
10
(kT NF W)
where the effective noise bandwidth W = 3.84 x 10
6
Hz,
and kT = 1.3804 x 10
23
x 290 W/Hz.
The noise figure at the UE is NF
[dB]
= 7 dB.
The macrocell antenna gain is calculated as
dB in level gain maximum 16dB
dB in level gain sidelobe 20dB
peak from down 3dB is pattern gain where angle 180 / 70
with , , 12 min ) (
max
2
max
=
=
=

(
(

|
|

\
|
=
G
G
G G dB G
s
s


Lounge
Study
Kitchen U
t
i
l
i
t
y

T
o
i
l
e
t

Corridor
S
i
d
e
w
a
l
k





and study being the rooms that are visited the most. At each
room, the user will spend a certain amount of time at a
waypoint.
The amount of time the user spends at the waypoint is
normally distributed, as shown in Fig. 6. The probabilities of
the user being in a room, and the mean time spent in the room
were derived empirically from a short survey on user
behaviour during evening peak times (i.e. the period starting
from when the user comes home from work to before going to
sleep).
To simulate the effect of underlay macrocell users, i.e.
passersby, an outdoor user model is used. The outdoor model
simply moves the users along the sidewalk outside the house,
whose location is shown in Fig. 5. The users walk past the
front of the house from one end of the sidewalk to the other in
a straight line at a speed of 1ms
-1
. The entry point of the user
(from the left or right side of the house) is randomly generated,
as is the part of the sidewalk the user walks on. The inter-
arrival times of pedestrians in a residential area was
determined to be exponentially distributed using a separate
large-scale outdoor pedestrian simulation. As such, the inter-
arrival times used in the model are generated using an
exponential distribution, with different means used for
different scenarios.
B. Simulation assumptions
System level simulations were performed to derive the
number of the mobility events for passing and indoor users as
a function of the distance from the femtocell. A scenario with
7 macrocells with 3 sectors each is considered. Femtocells are
deployed randomly within the coverage area and re-use the
same frequency as the macrocells in a hierarchical cell
structure. An overview of the simulated scenario with different
macrocell range and power can be found in [11]. Key
simulation parameters including the propagation models for
path-loss, shadow fading, and the antenna gain for the
macrocell sectors assumed are shown in Table I. The
macrocell downlink transmit power assumed is 20W per sector
and 10% of the total power is allocated to pilot channels. The
cell range is calculated such that the received signal to noise
ratio inside of a house at the cell edge is 10 dB, full coverage
of the area, including indoor coverage at the cell assuming an
additional wall loss of 15 dB. This results in a maximum range
in direction of the highest sector gain of 1046 metres. As a
result, the obtainable throughputs from the macrocell are
mainly interference limited.
The femtocell pilot powers are configured and optimized as
described in Section III. It is assumed that a mobility event is
triggered when the received pilot signal of a new cell is 4 dB
higher than the pilot signal of the current cell for a time of 500
ms. When a mobility event is triggered, the duration of the
procedure is assumed to be 650 ms.
Fig. 7 illustrates one example of the simulated area around
one house and shows the received pilot power levels within the
coverage areas for both macrocell and femtocell where the
power is auto-configured using the distance based method 2
with a target cell radius of 10m.
C. Simulation results
Fig. 8 shows a comparison of the resulting average number
of mobility events per hour for the three auto-configuration
schemes alone and the three proposed self-optimization
methods. It is shown that using a fixed power results in the
highest number of mobility events. The distance based auto-
configuration method achieves a similar performance as the
simpler measurement based method. In this example with a
mean inter arrival time of 600s for passing users, all mobility
event based self-optimization methods perform equally well
and achieve the optimum performance. It is clearly shown that
the mobility event based self-optimization methods are able to


Fig. 8. Mobility events per hour with outdoor user mean inter
arrival time of 600s


Fig. 9. Indoor coverage with outdoor user mean inter arrival
time of 600s




Fig. 8. Pilot power CDF with outdoor user mean inter arrival
time of 600s


Fig. 9. Mobility events per hour with outdoor user mean inter
arrival time of 3600s

significantly outperform the other methods that do not take
mobility events into account.
Fig. 9 shows the resulting indoor coverage for the
investigated auto-configuration and self-optimization schemes.
Auto-configuration methods 2 and 3 achieve very similar
coverage. Using a static pilot power achieves on average a
higher coverage, since the power was selected such that only in
10% of all cases the target coverage is not achieved. This also
means that in 90% of all cases a higher than target range is
achieved resulting in the increased mean coverage shown with
the associated drawback, the high number of resulting mobility
events per hour. The self-optimizing methods achieve the
highest coverage when femtocells are deployed in good
locations. A lower coverage is achieved compared to the auto-
configuration methods for femtocells deployed in unsuitable
locations close to the footpath, which is a trade-off resulting
from the reduction in core network signalling which is of
higher priority. Also the lower performance in unsuitable
locations is not necessarily a disadvantage since it would
encourage the user to re-deploy the femtocell, which will result
in both better coverage for the user and a lower number of
mobility events.
Fig. 10 shows CDFs of the resulting pilot power for all
considered auto-configuration and self-optimization methods.
In general all methods result in a large variation of pilot
powers that depend on both the distance from the macrocell
and the deployment location within the house. It is shown that
the mobility event based self-optimization methods show the
largest variation in power. While in most cases they use higher
power than the other auto-configuration methods in order to
maximise the indoor coverage for the given environment,
outdoor mobility and deployment location, they also reduce
the pilot power in some cases to -45dBm. This happens to
prevent excessive mobility events in cases where the user
deploys the femtocell in an unsuitable location, for example
close to the window near the sidewalk. Current cellular
standards such as UMTS have not been designed for
femtocells and therefore specify much higher values for the
minimum pilot power. This issue is currently under discussion
for LTE, where a reduction of the minimum pilot value will be
required for the Home eNodeB.
Fig. 11 shows a second example of the average number of
mobility events per hour for the considered auto-configuration
and self-optimization methods where a lower inter arrival time
for outdoor users of 3600s is considered. Again the proposed
mobility event based self-optimization methods clearly
outperform all methods that do not take mobility events into
account. In this case not all self-optimization methods achieve
the optimal performance. Method 1 cannot achieve the
optimum performance when the femtocell is deployed close to
the sidewalk. Method 2 improves the performance slightly by
in many cases cannot converge to the global optimum for these
deployment locations. However, it might still be preferable to
use the sub-optimum self-optimization methods since they
encourage the user to re-deploy the femtocell in a better
location which will result in both better coverage and a lower
number of mobility events compared to the optimum solution
for a poor femtocell location.
V. CONCLUSION
In this paper, different auto-configuration and self-
optimization methods were investigated that aim to optimize
the coverage of femtocells. Novel coverage adaptation
methods were proposed that use information on mobility
events of passing and indoor users to optimize the femtocell
coverage in order to minimize the increase of core network
mobility signalling. It was shown that mobility event based
self-optimization of coverage can both significantly reduce the
total number of mobility events caused by femtocell
deployments and improve the indoor coverage for femtocells



deployed in suitable locations compared to simpler methods
that aim to achieve a constant cell radius. Even for femtocells
located in unsuitable locations the self-optimization methods
are preferable from an operators point of view since they
encourage a re-deployment to a better location, resulting in an
improved coverage for the end user and in a reduced number
of mobility events in the core network.
REFERENCES
[1] S. Carlaw, C. Wheelock, Femtocell market challenges and
opportunities, ABI research, research report, 2007.
[2] M. Bauer, P. Bosch, N. Khrais, L. G. Samuel, and P Schefczik,
The UMTS Base Station Router, Bell Labs Technical Journal,
vol. 11, no. 4, pp. 93 111, 2007.
[3] D. Das, T. E. Klein, K. K. Leung, S. Mukherjee, G. Rittenhouse,
L. G. Samuel, H. Viswanathan, H. Zheng, Distributed radio
link control protocol facilitating fast cell site selection in an all-
IP cellular network, in Proc. Vehicular Technology Conference
(VTC), pp. 1692 1696 , 2005.
[4] S. Das, T. E. Klein, S. Mukherjee, G. Rittenhouse, L.G. Samuel,
H. Viswanathan, K. K. Leung, H. Zheng, Distributed Paging
and Registration in Wireless Networks, IEEE Network
magazine, Sept. 2005.
[5] L. T. W. Ho, L. Samuel, J. Pitts, Decentralised Cell
Dimensioning in 4G Networks: Analysing Self-Organising
Behaviour Using Complexity Metrics", World
Telecommunications Congress, 2002.
[6] L. T. W. Ho, L. Samuel, J. Pitts, "Applying Emergent Self-
organizing Behaviour for the Coordination of 4G Networks
Using Complexity Metrics", Bell Labs Technical Journal, vol.
8, no. 1, pp.5-25, 2003.
[7] L. T. W. Ho, F. J. Mullany, H. Claussen, L. G. Samuel,
Autonomous organization of wireless network transport in a
multi-provider environment, in Proc. International Symposium
on Autonomous Decentralized Systems (ISADS), 2005.


























[8] F. J. Mullany, L. T. W. Ho, L. G. Samuel, H. Claussen, Self-
deployment, self-configuration: Critical future paradigms for
wireless access networks, in Lecture Notes in Computer
Science (LNCS), Springer Verlag, vol. 3457, pp. 58-68, 2005.
[9] H. Claussen, Distributed Algorithms for Robust Self-
Deployment and Self-Configuration in Autonomous Wireless
Access Networks, in Proc. IEEE International Conference on
Communications (ICC), 2006.
[10] H. Claussen, Autonomous self-deployment of wireless access
networks in an airport environment, in Lecture Notes on
Computer Science (LNCS), Springer Verlag, vol. 3854, pp. 86-
98, 2006.
[11] H. Claussen, Performance of macro- and co-channel femtocells
in a hierarchical cell structure, in Proc. IEEE International
Symposium on Personal, Indoor and Mobile Radio
Communications (PIMRC), 2007.
[12] L. T. W. Ho, H. Claussen, Effects of user-deployed, co-channel
femtocells on the call drop probability in a residential scenario,
in Proc. IEEE International Symposium on Personal, Indoor
and Mobile Radio Communications (PIMRC), 2007.
[13] H. Claussen, L. T. W. Ho, H. R. Karimi, F. J. Mullany, L. G.
Samuel I, Base Station Cognisant robots and future wireless
access networks, in Proc. Consumer Communications and
Networking Conference (CCNC), pp.595 599, 2006.
[14] H. R. Karimi, L. T. W. Ho, H. Claussen, L.G. Samuel,
Evolution Towards Dynamic Spectrum Sharing in Mobile
Communications, in Proc. IEEE International Symposium on
Personal, Indoor and Mobile Communications (PIMRC), 2006.
[15] H. Claussen, L. T. W. Ho, L.G. Samuel, Financial Analysis of a
Pico-Cellular Home Network Deployment, in Proc.
International Conference on Communications (ICC), 2007.
[16] 3GPP TR 23.882 3GPP system architecture evolution: report
on technical options and conclusions, Section D2.4, v1.11.0,
2007.
[17] W. Webb, Wireless Communications: The Future, John Wiley
& Sons, 2007.
[18] C. Anderson, The Long Tail: Why the Future of Business Is
Selling Less of More. New York, Hyperion, 2006.

Anda mungkin juga menyukai