Content uploaded by Yu Peng
Author content
All content in this area was uploaded by Yu Peng on Apr 20, 2015
Content may be subject to copyright.
A Distributed Cell Outage Compensation
Mechanism Based on RS Power Adjustment in
LTE Networks
LI Wenjing, YU Peng, YIN Mengjun, MENG Luoming
State Key Laboratory of Networking and Switching Technology, Beijing University of Posts and Telecommunications,
Beijing 100876
China
Abstract: To solve the coverage and quality
problems caused by cell outage in LTE
networks, this paper proposes a distributed self-
organizing networks management architecture
and a distributed cell outage compensation
management mechanism. After detecting and
analyzing the outage, a cell outage
compensation algorithm based on reference
signal power adjustment is proposed. The
simulation results show that the proposed
mechanism can mitigate the performance
degradation significantly. Compared with other
algorithms, the proposed scheme is more
effective in compensating the coverage gap
induced by cell outage.
Keywords: LTE; cell outage compensation
(COC); reference signal power; PSO
I. INTRODUCTION
Self-healing is one of the functions in the Self-
Organizing Networks (SON), which has become
one of the hottest topics in telecommunication
network research field [1]. The purpose of self-
healing is to automatically detect and locate
faults, meanwhile, resolve the possible loss of
coverage and capacity. An important aspect of
self-healing is cell outage management (COM)
which comprises two functions: cell outage
detection (COD) and cell outage compensation
(COC) [2]. This paper will concentrate on the
COC.
Cell outage is the total loss of radio services
in the coverage area of a cell because of some
internal or external failures [3]. Cell outages
originate from different reasons, e.g., hardware
and software failures, external failures, or
erroneous configuration. The goal of COC is to
mitigate the degradation of coverage, capacity
and service quality caused by the cell outage. In
the COC procedure, to satisfy the operator’s
performance requirements, network parameters
will be adjusted automatically based on the
coverage objectives and other quality indicators
without human interference [4].
II. RELATED WORKS
Recently, many organizations and projects, such
as 3rd Generation Partnership Project (3GPP),
Next Generation Mobile Networks (NGMN)
and Self-Optimization and self-ConfiguRATion
in wirelEss networkS (SOCRATES) project, all
conduct research on self-healing. 3GPP gives a
detailed description of self-healing including
concepts, requirements, use cases, background
information and processes [3]. NGMN research
report gives analysis on control parameters of
COC, the variable influence on COC for
different scenarios, and the effect evaluations
from different measurement data [5].
SOCREATES project also describes the
procedure, scenarios and framework of self-
healing, and indicates that the coverage gap
could be compensated by adjusting the
parameters, such as channel parameters (e.g.,
reference signal transmit power or the uplink
target received power) and the antenna
parameters (e.g., antenna downtilt) of the
surrounding cells to achieve COC [6].
The proposals and reports mentioned above
are instructive to the research of COC, but all of
them are not giving out the specific mechanisms
and algorithms for COC.
Some specific COC algorithms are put
forward to achieve the compensation for the
outage area by adjusting the pilot power, the
uplink target received power and other
parameters of surrounding cells [7][8]. But all
these algorithms are adjusting parameters step
by step. Moreover, for each adjustment, the data
must be re-collected and assessed, thus, the
outage status cannot be compensated in time. In
paper [9], a centralized COC mechanism and a
heuristic COC algorithm are proposed to adjust
the power of neighbor eNodeBs. However, the
efficiency of the centralized COC mechanism is
relatively low. At the same time, the
mathematical model proposed by [9] is mainly
based on the information collected from users,
and when users under outage cell are few,
compensation outcomes may not be acceptable.
This paper puts forward a distributed COC
mechanism based on reference signal power
adjustment to fix the coverage gap and other
problems caused by the sudden outage in the
LTE network. The problems such as low speed
of constringency, low efficiency, depended too
much on user information, and incomplete
mathematical model in the above papers will be
solved by our COC scheme.
III. COC MANAGEMENT ARCHITECTURE
3.1 SON management architecture
In general, a SON entity is a logical automatic
management entity which can implement one or
more SON functions without the human
interference. To fulfill the cooperation among
SON functions, the SON coordinator is needed.
According to the location of SON entities,
SON management architecture can be classified
as centralized, distributed and hybrid
architectures [10].
In the centralized management architecture
which is adopted in paper [9], the SON entity is
located in a management center (such as OAM).
This architecture is simple and easy to
implement. But in the flattening LTE/LTE-A
network environment, this architecture has its
shortcomings. 1) OAM needs communicate
frequently with eNodeBs through management
interface. For the real-time information, the
efficiency of the management interface is
relatively low. 2) As a centralized control node,
when OAM failed, the whole SON system will
be unavailable. 3) The OAM will limit the
performance and scalability of the system, and
the OAM becomes the bottle-neck of the whole
SON system.
In the distributed management architecture,
the SON entities are located in the network
elements, that is to say the autonomic
management control loop[9] (called as AMCL
in this paper) is implemented in the network
elements. Information is exchanged among
eNodeBs through X2 interface when necessary,
shown in the Fig.1.
eNodeB
eNodeB
X2
OAM
SON
Management
Strategy
Management
Interface
Management
Interface
SON Entity
SON
Function
AMCL
SON
Function
AMCL
SON
Function
AMCL
SON Coordinator
Control
Parameters
Measurement
Data
Measurement
Data
Control
Parameters
SON Entity
SON
Function SON
Function
AMCL
SON
Function
AMCL
SON Coordinator
AMCL
Fig.1 Distributed management architecture
In this architecture, the management interface
between OAM and network elements is used to
transmit the SON management strategies and
other necessary information. OAM will not
control the execution of the autonomic
management control loop (AMCL).
Compared to the centralized management
architecture, the distributed architecture can
enhance the management efficiency, and this
architecture is effective to avoid the fatal
consequences when OAM fails. At the same
time, because the information exchanged among
network elements may be conflicting, the
conflict coordination mechanism must be
established, while signaling overhead should be
limited within the allowed scope.
In this paper, a distributed COC management
mechanism is proposed.
3.2 COC management mechanism
In the distributed COC management architecture,
the main function of OAM is to generate and
maintain the COC strategies based on the
operator-formulated compensation policy and
the network environments. The process of the
distributed COC management mechanism is
illustrated in Fig.2.
OAM eNodeB eNodeB-Co eNodeB
3) SON Strategy (specify eNodeB-Co)
1) Monitor Conditions of Outage
2) Outage Detected
3) SON Strategy (specify eNodeB-Co)
3) SON Strategy (specify eNodeB-Co)
4) necessary info 4) necessary info
5) Evaluate the coverage problem
6) Generate an optimal
compensation scheme
7) parameter modify request
7) parameter modify request
8) Adjust relevant parameters of the corresponding cells
9) necessary info 9) necessary info
10) Evaluate the compensation result
10) Report the compensation result
Monitoring
Analyzing
Planning
Executing
Evaluating
Fig.2 Process of distributed COC mechanism
The steps are described as follows.
Phase 1. Monitoring Phase.
1) In this phase, the OAM and the eNodeBs are
all monitoring the corresponding information,
such as Key Performance Indicators (KPIs),
statistical data, alarms, Radio Link Failure
(RLF) counters, and other measurement data.
In this phase, they all monitor the conditions
of the outage.
2) According to the data collected above, OAM
can decide whether outage occurs or not. If no
outage is detected, continue monitoring.
3) If outage situation is detected, OAM notifies
the eNodeBs surrounding the outage cell that
the outage is detected, and sends COC
management strategy to these eNodeBs. For
example, OAM can specify a coordinating
eNodeB (called eNodeB-Co) among the
surrounding eNodeBs to play a role of
coordinator of the outage area. In this paper,
eNodeB-Co will implement the analyzing and
planning functions and deal with the possible
conflict during COC process.
Phase 2. Analyzing Phase.
4) At the beginning of analyzing phase, eNodeB-
Co will collect more data from other eNodeBs
through X2 interface, including geographical
coordinates, transmission power, coverage
area, antenna parameters, payload situation,
etc.
5) ENodeB-Co analyzes data from previous steps
and evaluates the coverage problems and
coverage objectives. ENodeB-Co determines
whether compensation is needed. If the
problem cannot be recovered directly, go to 6)
to generate a COC scheme..
Phase 3. Planning Phase.
6) In this phase, based on the data collected in
the above phases, eNodeB-Co will generate a
specific optimal compensation scheme using
the algorithm proposed in section IV.
Phase 4. Executing Phase.
7) After compensation scheme is generated,
eNodeB-Co sends the parameter adjustment
requests through X2 interface to the related
eNodeBs to extend their coverage area.
8) The corresponding eNodeBs adjust relevant
parameters according to the requests.
Phase 5. Evaluating Phase.
9) After COC, it is necessary to evaluate the
performance of compensation and assess
whether the network returns to the normal
state. The eNodeBs send necessary
information to eNodeB-Co through X2
interface.
10) ENodeB-Co evaluates the compensation result
and reports the COC result to OAM, then
monitoring phase is entered again.
IV. COC ALGORITHM
In LTE network, reference signal is mainly used
for channel estimation, which can be used by the
users as the reference symbol to perform the
downlink measurement, synchronization and data
demodulation. Increasing the power of RS (PRS)
can effectively increase the coverage radius. This
section proposes an algorithm to achieve COC by
adjusting PRS.
4.1 Model of COC
Suppose the eNodeB corresponding to outage
cell is eNB0, and the surrounding eNodeBs are
denoted as N={eNB1, eNB2,…, eNBn}, of which
the total number is n. the PRS corresponding to N
are denoted as P={p1RS, p2RS,…, pnRS}. S(piRS) is
the coverage area of eNBi related to piRS. The
coverage area of all eNodeBs is denoted as Starget.
Here is the calculation method of S(piRS).
RSRP is a key parameter to evaluate the downlink
coverage. In order to fulfill the requirements of
effective coverage, the received RSRP of cell-
edge users should be stronger than a minimum
value. Define the minimum RSRP as γ, the
maximum allowable path loss PLimax can be
calculated by [11]:
max RS
ii
PL p
(1)
After the maximum allowable path loss is
known, the maximum coverage radius Ri of the
cell can be calculated as follows [12]:
1max
(( _ )| , , )
i
i BS MS
R g PL sh margin f h h
(2)
Where, sh_margin is the shadow fading margin,
f is the frequency of the system, hBS and hMS are
the antenna heights of eNodeB and UE
respectively. The function g() is determined by
adopted path loss model, e.g., the Cost231-Hata
model or the Okumura model. The quantities
before “|” in (2) are variables and varies in a
continuous interval, while the quantities after “|”
are parameters with discrete known values. After
Ri is calculated, combine (1) and (2), then the
relationship between S(piRS) and piRS can be
expressed as follows:
12
max
( ) ( ( ))
ii
RS
S p g PL
(3)
There are two optimization objectives in COC
algorithm. The first optimization objective is to
maximum the coverage rate of whole region fcov,
expressed as follows:
12
cov ( ) ( ) ( )
100 n
RS RS RS
target
S p S p S p
fS
K
(4)
Raising the PRS of the surrounding cells can
effectively increase the service area, covering
whole or parts of the outage area. However, when
the service area enlarges, the excessive coverage
overlap tends to appear, and pilot pollution may
arise as well. For simplicity, in this paper, we only
consider triple overlap. To reduce the coverage
abnormal, the second optimization objective is to
minimum the overlap rate foverlap, expressed as
follows:
1 , , , ( ) ( ) ( )
nn i j k
RS RS RS
i j k N j k i
overlap target
S p S p S p
fS
(5)
Meanwhile, a limited range of PRS is
considered. In LTE networks, PRS is varied from
27 to 39dBm, with a default setting of 33dBm [3].
Let PRSmax be the maximum of PRS, a constraint of
eNodeB is defined as:
max
i
RS RS
PP
(6)
According to the above discussion, the
problem of the algorithm is to optimize the two
objectives at the same time. To solve this problem
under condition (6), (4) and (5) are combined into
one global fitness function. In this way, both the
coverage and the overlap are taken into account:
cov cov 1
()
1
tot overlap
f f k f f
(7)
Where
cov
2
cov
cov cov
0,
( ) 100 ( ) ,
100,
f
f
k f f
else
(8)
Where, η and φ are two boundary values of
coverage rate, changing depending on the network
environment. According to the different value of
coverage rate and overlap rate, the value of ftot
changes in range [0,200]. The maximum value
200 is obtained when fcov equals to 100 and foverlap
equals to 0. If fcov is below boundary value η, that
means coverage is far worse than acquired, at this
time, to optimize the coverage is more important.
Therefore, a factor k(fcov) of 0 is chosen when
coverage is below η. In that way, the coverage
requirement can be quickly obtained. When fcov is
between η and φ, factor k(fcov) increases as fcov
increases. When fcov approaches φ, the coverage
requirement is almost fulfilled. The closer fcov is to
φ, the more important it becomes to start
optimizing foverlap in the network. When fcov is over
φ, k(fcov) should maintain maximum. In this way,
we use weight factor k(fcov) to balance two
optimizing objectives as well as optimize them at
the same time.
4.2 Algorithm description
In the phase of planning, COC algorithm will
generate the PRS adjustment scheme by getting the
largest ftot. Here we use Particle Swarm
Optimization (PSO) algorithm to solve the
problem. PSO uses a simple mechanism that
mimics swarm behavior in birds flocking to guide
the particles to search for globally optimal
solutions. As PSO is easy to implement, it can
perform a global search over the entire search
space with faster convergence speed and with
high accuracy [13].
In COC algorithm, the goal of finding the
solution set of PRS adjustment scheme in
surrounding cells is converted into finding the
optimal positions of particles. The particle number
is m, and each particle i is associated with two
vectors, where position vector denoted as Xi = ( xi1,
xi2,…, xiD ), representing the values of PRS, and
velocity vector denoted as Vi = ( vi1, vi2,…, viD ),
representing the changes of PRS. D stands for the
dimensions of the solution space, determined by
the number of eNodeBs involved in compensation.
In each iteration, particles calculate the value of
fcov and foverlap according to the current position,
then using (7) and (8) to get the result of global
fitness function ftot. The next step is updating the
best position of each particle which is denoted as
Pi = (pi1, pi2, …, piD ), and the previous global best
position denoted as Pg = ( pg1, pg2, …, pgD )
according to ftot.
Before finding the best position, the velocity
and position of each particle i are updated
according to the following formula [14]:
11 1 2 2
( ) ( )
t t t t t t
id id id id gd id
v v c r p x c r p x
(9)
11t t t
id id id
x x v
(10)
Where c1 and c2 are accelerating factors, r1 and
r2 are two uniformly distributed random numbers
between [0, 1], ω is inertia weight factor. The
process of COC algorithm is described in Fig. 3.
Start, t=0
Initialization,Initialize number
of particles, initial position and
velocity, Tmax
Update Vi according to
Equation (9)
Update Xi according to
Equation (10)
t = t+1
t < Tmax
Export the solution set
no
yes
EvAluate the fitness value ftot,
Find the current Pi and Pg
Fig.3 Process of COC algorithm
V. SIMULATION AND DISCUSSION
In this section, we analyze the performance of the
COC algorithm in two parts: coverage gain
analysis and comparison with other algorithm.
5.1 Simulation environment
The simulation is based on LTE network, in an
urban area with 4.5km × 4.5km. According to the
trial test and scale technology test of LTE in
China, we set a typical urban area with 7 eNodeBs,
ranging from 300m to 650m, as the experiment
simulation environment. 200 UEs are randomly
distributed within the range. The key parameters
concerned in simulation are shown in Table I. The
parameters of PSO are set in Table II.
Table I Simulation parameters
Parameter
Values
System bandwidth
20 MHz
Channel mode
Typical urban
Carrier frequency
2600 MHz
Uplink transmit power
23 dBm
Downlink transmit power
46 dBm
Antenna mode
3GPP 3D Model
Antenna height
35m(eNB),1.5m(UE)
Antenna downtilt
15°
Path loss model
Cost231-Hata
Shadowing
8dB
Penetration loss
20dB
Cable loss
2dB
Gain of eNB antenna
18dBi
Table II Parameters of PSO
Parameter
Values
m
30
c1, c2
2
r1, r2
Random in[0, 1]
ω
η
0.8
90
φ
95
PRSmax
39 dBm
Tmax
200
5.2 Simulation result and discussion
1) Coverage gain analysis
Through the adjustment of PRS, the coverage
areas of surrounding cells are increased in
simulation. ENodeB0 is located in the middle of
the region and is selected to be the outage cell.
We can get a set of PRS adjustment value shown in
Table III. It can be seen that the PRS of all the six
surrounding cells are adjusted according to COC
algorithm. Cell 5 is set to the maximum value.
Table III Value of PRS
Cell No.
Initial/dBm
After COC/dBm
0
33
Outage
1
33
37.93
2
33
38.52
3
33
36.46
4
5
6
33
33
33
36.32
39
34.19
Fig.4 is the RSRP intensity distribution of
whole region before and after COC. Because the
outage of eNodeB0 is in the middle, RSRP of
middle area is extremely low in Fig.4(a), the
coverage gap inevitably appears. After COC in
Fig. 4(b), the adjustment of PRS makes the RSRP
improved significantly in the middle area, which
means the scheme is effective in compensating the
coverage gap.
During the simulation process, the changes of
coverage and overlap rate are shown in Fig.5(a).
The simulation results show that by adjusting
PRS of surrounding cells, the coverage rate can
reach to 98.5%, meanwhile the overlap rate is
about 1.6%. In actual LTE network, when
coverage rate reaches 98%, it can be considered as
effective coverage; thus the compensation results
have achieved the goal of coverage objective.
Meanwhile, the Fig.5(a) also shows that the
algorithm is convergent.
During the process of COC algorithm, the
changes of ftot are shown in Fig. 5(b).
0 4.5
0
4.5
-120
-110
-100
-90
-80
-70
-60
NO.1
NO.6
NO.5
NO.4
NO.3
NO.2
0 4.5
0
4.5
-120
-110
-100
-90
-80
-70
-60
NO.2
NO.4
NO.3
NO.1
NO.6
NO.5
(a) Before COC (b) After COC
Fig.4 RSRP distribution before and after COC
020 40 60 80 100 120 140 160 180 200
50
55
60
65
70
75
80
85
90
95
100
Number of Iterations
020 40 60 80 100 120 140 160 180 200
0
0.05
0.1
0.15
coverage
overlap
020 40 60 80 100 120 140 160 180 200
40
60
80
100
120
140
160
180
200
Number of Iterations
Value of function
(a) Changes of fcov and foverlap (b) Changes of ftot
Fig.5 Changes of fcov , foverlap and ftot
At the beginning of the process, the value of
fitness function ftot is low due to the low coverage.
Meanwhile, vibration amplitude presents great
changes as the value of k is not stable. When fcov
stabilized at a higher value, the fitness function ftot
has gradually stabilized and gradually moved to
the optimal value with the increase of the number
of iterations. When ftot converges to the maximum,
it is the optimal compensation scheme. A
reasonable trade-off has been achieved between
two optimization targets, coverage rate and
overlap rate.
2) Algorithm performance comparison
In paper [9], a COC algorithm named
Autonomic Particle Swarm Compensation
Algorithm (APSCA) is proposed. APSCA can
implement the compensation effectively in most
cases, but the model which depends on the
information reported by users has some limits. If
the distribution of users is not even or the users
are few, the compensation result may not ideal.
The algorithm proposed by this paper can achieve
effective compensation without dependency on
the information reported by users.
Fig. 6 shows the cumulative probability
distributions of received RSRP and SINR in three
cases. According to the effective coverage
reference value of the LTE network, the
cumulative probability of RSRP>−105dBm is
more than 95%. Combing with the coverage
reference value, as illustrated in Fig.6(a), before
COC, the cumulative probability of RSRP being
less than −105dBm equals to 13.64%, which is far
worse than the effective coverage requirement.
After COC, the cumulative probability of RSRP
being more than −105dBm equals to 95.4%,
which is within the scope of effective coverage
requirement. The simulation result of APSCA is
also shown in Fig. 6(a). The cumulative
probability of RSRP being more than −105dBm
equals to 96.1%, almost the same as the result of
COC algorithm in this paper, but observing the
overall distribution of RSRP, the COC algorithm
in this paper ensures stronger signal distribution.
As illustrated in Fig.6(b), before COC, due to
the reducing of received power as well as the
increasing of interference, cell outage leads to a
low level of user SINR in the region. When
compensation is performed, both the algorithm in
this paper and APSCA are effective to improve
the SINR. Since minimizing the overlap is one of
the optimization objectives, it is successful in
preventing the interference.
Fig.6 Cumulative probability of RSRP and SINR
By comparing the average user throughput and
the edge-user (the 5th percent downlink channel
quality) throughput in Fig.7, our algorithm is
effective in increasing the average throughput,
and the performance is better than APSCA. As
seen, the increasing of edge-user throughput is
more significantly than that of the average
throughput. This is an additional trade-off that
needs to be considered in the future.
0
0.05
0.1
0.15
0.2
0.25
0.3
edge-average
throughput (Mbps)
after compensation
APSCA
before compensation
0
0.5
1
1.5
2
2.5
3
3.5
regional-average
throughput (Mbps)
after compensation
APSCA
before compensation
Fig.7 Throughput comparison
According to the above analysis of simulation
results, the COC algorithm proposed in this paper
is effective to achieve outage compensation,
ensuring regional coverage at a higher level, and
reducing the emergence of coverage overlap.
Meanwhile the performance is kept on the
accepted levels guaranteeing the quality of
services.
VI. CONCLUSION
This paper has proposed a distributed COC
management mechanism based on RS power
adjustment. The simulation results prove that the
-110 -100 -90 -80 -70 -60 -50 -40
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
RSRP (dBm)
C.D.F.
after Compensation
before Compensation
APSCA
-20 -15 -10 -5 0 5 10 15 20 25 30
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
SINR (dB)
C.D.F.
after Compensation
APSCA
before Compensation
(a) RSRP (b) SINR
mechanism can effectively solve the coverage
problems caused by COC in LTE network.
In future research, more control parameters
will be discussed, such as antenna downtilt, uplink
target received power, scheduling parameters etc.,
or combination of those parameters. Meanwhile,
we will continue our work on more efficient
methods using the Coordinated Multiple Points
Transmission and Reception (CoMP) technology
in the COC.
ACKNOWLEDGMENT
This research is supported by NSFC (61271187)
and 863 Project (No. 2014AA01A701)
REFERENCES
[1] J.Ramiro and K. Hamied. Self-organizing
networks: self-planning, self-optimization
and self-healing for GSM, UMTS and LTE.
John Wiley & Sons Ltd, 2012.
[2] M. Amirijoo, L. Jorguseski, T. Kürner, R.
Litjens, M. Neuland, L.C. Schmelz and U.
Türke. “Cell Outage Management in LTE
Networks”, Proceedings of the 2009 6th
International Symposium on Wireless
Communication Systems, ISWCS'09, pp.
600-604, 2009.
[3] 3GPP TS 32.541 V10.0.0, Telecommunication
management; Self-organizing Networks
(SON); Self-healing concepts and
requirements, 2011.
[4] M. Amirijoo, L. Jorguseski, T. Kürner, R.
Litjens and R.Nascimento, “Effectiveness of
cell outage compensation in LTE networks”,
2011 IEEE Consumer Communications and
Networking Conference, CCNC'2011, pp.
642-647, 2011.
[5] Next Generation Mobile Networks
Recommendations on SON and OAM
Requirements, by the NGMN alliance, 2008.
[6] SOCRATES Deliverable D5.9: Final Report
on Self-Organisation and its Implications in
Wireless Access Networks. 2010.
[7] Fuqiang Li, Xuesong Qiu, Luoming Meng and
Heng Zhang, “Achieving cell outage
compensation in radio access network with
automatic network management”, 2011 IEEE
GLOBECOM Workshops, pp. 673-677, 2011.
[8] Fuqiang Li, Xuesong Qiu, Wenjing Li and H.
Wan, “High load cell outage compensation
method in TD-SCDMA wireless access
network”, Journal of Beijing University of
Posts and Telecommunications, v 35, n1, pp.
32-35+45, February 2012.
[9] Wenjing Li, Peng Yu, Zhengxin Jiang and
Zifan Li, “Centralized Management
Mechanism for Cell Outage Compensation in
LTE Networks”, International Journal of
Distributed Sensor Networks, v 2012.
[10] Mugen Peng, Dong Liang, Yao Wei and Jian
Li, “Self-Configuration and Self-
Optimization in LTE-Advances
Heterogeneous Networks”, IEEE
Communication Magazine, pp 36-45, May
2013.
[11] M. Amirijoo, L. Jorguseski, R. Litjens and
L.C.Schmelz, “Cell outage compensation in
LTE networks: Algorithms and performance
assessment”, 2011 IEEE 73rd Vehicular
Technology Conference, VTC2011-Spring,
2011.
[12] L.Song and J. Shen. Evolved Cellular
Network Planning and Optimization for
UMTS and LTE, CRC Press, 2011.
[13] I.C. Trelea, “The particle swarm optimization
algorithm: Convergence analysis and
parameter selection,” Inf. Process. Lett. , vol.
85, no. 6, pp. 317-325, March 2003.
[14] Z. Zhan, J. Zhang, Y. Li and H. S. H. Chung,
“Adaptive particle swarm optimization”,
Systems, Man, and Cybernetics, Part B:
Cybernetics, IEEE Transactions on Volume:
39, Issue: 6, pp.1362-1391, 2009
LI Wenjing, July 1973, Associate Professor of
BUPT, M.S. advisor. Her research interests
include SON, wireless network management and
future network management.