Hindawi

Mobile Information Systems

Volume 2017, Article ID 4380676, 12 pages
https://doi.org/10.1155/2017/4380676

Research Article
Self-Planning of Base Station Transmit Power for
Coverage and Capacity Optimization in LTE

V. Buenestado,1 M. Toril,1 S. Luna-Ramírez,1 and J. M. Ruiz-Avilés2

1
Department of Communication Engineering, University of Málaga, 29071 Málaga, Spain
2
Ericsson, 29590 Málaga, Spain

Correspondence should be addressed to V. Buenestado; vbg@ic.uma.es

Received 7 April 2017; Accepted 11 July 2017; Published 22 August 2017

Academic Editor: Antonio de la Oliva

Copyright © 2017 V. Buenestado et al. This is an open access article distributed under the Creative Commons Attribution License,
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

A computationally efficient self-planning algorithm for adjusting base station transmit power in a LTE system on a cell-by-cell
basis is presented. The aim of the algorithm is to improve the overall network spectral efficiency in the downlink by reducing
the transmit power of specific cells to eliminate interference problems. The main driver of the algorithm is a new indicator that
predicts the impact of changes in the transmit power of individual cells on the overall network Signal to Interference plus Noise
Ratio (SINR) for the downlink. Algorithm assessment is carried out over a static system-level simulator implementing a live LTE
network scenario. During assessment, the proposed algorithm is compared with a state-of-the-art self-planning algorithm based on
the modification of antenna tilt angles. Results show that the proposed algorithm can improve both network coverage and capacity
significantly compared to other automatic planning methods.

1. Introduction the most interfering ones, so that a wider coverage usually

leads to a higher interference, a lower spectral efficiency, and,
In recent years, mobile communications have experienced a ultimately, less capacity. As a result, network coverage and
rapid increase in the number of users and services, which capacity in LTE are strongly linked, so a tradeoff between
has led operators and manufacturers to develop systems them is necessary [3].
with greater capacity. In parallel, the complexity and size of Power control (PC) is a powerful strategy for dealing
these systems have increased exponentially, making network with the CCO problem. The aim of PC is to reduce the
management a very challenging task. To deal with such com- amount of interference from neighbor cells while ensuring
plexity, operators demand automatic tools for configuring that enough power is transmitted to (or received from) User
network parameters, as a flexible solution to improve network Equipment (UE) to maintain an acceptable link quality [4, 5].
capacity without new investments. This trend has stimulated In LTE, fractional power control is used in the UpLink (UL)
research and standardization activities in the field of Self- to dynamically change UE transmit power [5]. Moreover,
Organizing Networks (SON) [1]. several self-planning methods have been proposed to adapt
Network Coverage and Capacity Optimization (CCO) UpLink Power Control (ULPC) parameters in LTE to local
has been identified by operators as one of the most important network conditions [6–10]. However, for the downlink (DL)
use cases of SON [2]. The aim of CCO is to provide optimal of LTE, power planning is the simplest solution to solve CCO
(i.e., maximum) coverage and capacity. In legacy radio access issues in the absence of a power control scheme.
technologies (e.g., Global System for Mobile communication, Changing the base station transmit power is as costly
GSM), CCO can be solved easily as coverage and capacity as changing any other radio access network parameter since
are decoupled by means of frequency planning. This is not both actions only require modifying the network parameter
the case for Long Term Evolution (LTE) technology, where file. However, there remains the problem of finding the
all cells in the same layer use the same frequency band. This optimum transmit power settings. A tradeoff exists between
tight frequency reuse scheme makes the closest cells also ensuring a good connection quality for users served by the
2 Mobile Information Systems

cell changing its transmit power while reducing interference model but make use of measurements from the live network
in its neighbor cells. Thus, power planning can be formulated to adjust network parameters. Finally, self-healing methods
as a large-scale nonseparable multiobjective optimization aim to detect, diagnose, and compensate problems caused by
problem. To find the optimal power plan, many search abnormal events in the network [21–23].
algorithms have been proposed in the literature [11–16]. To SON methods can also be classified in terms of the
check the quality of a plan, analytical network models can be modified network parameter. In particular, CCO is usually
used [14]. However, analytic approaches fail to reflect relevant performed by changing antenna bearings [6, 19, 20, 24–
network conditions (irregular spatial traffic distribution, base 30] or power settings [11–13, 15, 16]. Strategies to modify
station configuration, channel conditions, clutter type, etc.). antenna bearings include both self-planning [6, 19, 24–
To solve this limitation, a system-level simulator is often used 26] and self-optimization methods [20, 27–30]. Although
in network planning tools [11–13, 15, 16]. However, system- most of these methods modify antenna tilt angles, a few
level simulations are time consuming. Thus, it is essential of them simultaneously set tilt and azimuth angles [6, 24,
to have an automatic search algorithm that finds the best 26]. Regarding power-based CCO methods, a power control
parameter plan in a few attempts. To the authors’ knowledge, algorithm is proposed in [11] for adjusting power levels
no previous work has derived a simple rule to modify an according to the needs of UEs. The algorithm tunes DL
existing power plan ensuring that the resulting solution is transmit power based on signal quality measurements from
indeed the best solution. UEs to ensure that all UEs experience adequate transmission
In this work, a novel self-planning algorithm for modi- quality. The algorithm also detects when UEs experience a
fying DL transmit power in a LTE system is presented. The transmission quality greater than required, so that transmit
proposed algorithm aims to improve the overall network power is decreased to minimize interference in neighbor
SINR by adjusting DL transmit power on a per cell basis. cells. In [12], the previous algorithm is extended by solv-
The algorithm is designed as a set of controllers (one per ing convergence problems through starting planning with
cell), whose input is a new cell performance indicator. This maximum cell power levels for all transmitters. In [13], a
novel indicator, which is the core of the proposal, reflects if a decentralized algorithm for adjusting the transmit power of
higher (or lower) transmit power in a particular base station base stations in LTE is also presented. In this case, each
increases (or decreases) the overall network SINR in DL. neighbor cell searches for the optimal setting of the transmit
The algorithm is validated on a static system-level simulator power of the cell, keeping the power setting of the rest
implementing a real LTE dense urban scenario. During the of cells unaltered. The algorithm relies on the knowledge
analysis, the proposed algorithm is compared with a state-of- of the interference produced by a base station in its neighbor
the-art self-planning approach based on remote electrical tilt. cells at different transmit power levels. In [15], a more
The rest of the work is organized as follows. Section 2 sophisticated algorithm is proposed to jointly tune transmit
presents the state of research in the CCO problem. Section 3 power and antenna tilt. In this algorithm, cells are classified
presents the problem formulation from which the new in three groups depending on the ratio of covered UEs
indicator is derived. Section 4 outlines the proposed method and carried traffic. Parameter changes are calculated based
for adjusting DL transmit power on a cell basis. Section 5 on the performance of the worst cell, and changes are
shows the results of simulations carried out to validate the simultaneously executed for all cells in the same group.
algorithm. Finally, Section 6 presents the main conclusions Therefore, this method does not exploit the fact that network
of the work. parameters can be set on a cell-by-cell basis. In [16], a
decentralized self-optimization algorithm is proposed to
2. Related Work adjust both transmit power and tilt on a cell basis to
maximize both cell-average and cell-edge UE throughput
Depending on their purpose, SON methods can be classified of the whole network. The algorithm is implemented as
as self-planning, self-optimization, or self-healing [17, 18]. a controller based on fuzzy reinforcement learning. Inputs
Self-planning methods are conceived for the planning stage, are the current transmit power and tilt settings of the
when the network is not deployed yet. Consequently, self- optimized cell, and the average relative differences in load and
planning algorithms make use of network models to estimate spectral efficiency with neighbor cells. A central controller
the quality of a network parameter plan. In the search of enables cooperative learning by sharing the result of the
the best plan, many different parameter settings have to be adaptation process among cells. Controller parameters are
tested. For this purpose, classical optimization methods are adapted based on the result of random parameter changes.
used to find the best solution (e.g., brute-force enumeration A major drawback is the fact that the algorithm needs many
[19], simulated annealing [13], Taguchi [6], and genetic [20]). iterations to converge and can temporarily degrade network
To check the quality of a plan, analytical [14] or simulation performance.
models [5] can be used. Self-optimization (a.k.a. self-tuning) Likewise, power replanning is also used in self-healing
methods are designed to adapt network parameters to chang- algorithms to solve localized problems caused by network
ing network conditions during the operational stage [11– failures. Unlike self-optimization approaches, the aim of heal-
13, 15, 16]. Self-optimization algorithms usually consist of a ing is not to achieve optimal system performance, but to bring
controller that iteratively modifies network parameters based a faulty cell to acceptable service levels. The most common
on certain network performance indicators. Unlike self- application is Cell Outage Compensation (COC) [21–23].
planning, self-optimization algorithms do not need a system The main limitation of these methods is the assumption
Mobile Information Systems 3

that, in normal operation, base stations do not fully utilize quality of a network, depending on how the performance of
the available transmit power. However, in live networks, users and cells is aggregated:
base stations are generally set to the maximum power to
provide maximum coverage. Moreover, these algorithms are (a) Overall user mean, considering all users equally. Thus,
mainly focused on improving coverage, with less emphasis on more populated cells tend to dominate the figure for
intercell interference or signal quality. merit.
A major drawback of most SON algorithms is the lack of (b) Cell arithmetic mean, where all cells are treated the
an optimality proof. Most self-planning methods are based same, regardless of their size and traffic. Such a figure
on heuristic approaches that reduce the size of the solution of merit is simple to compute and easy to interpret,
space to be explored to reduce computation time. Likewise, being the preferred option for operators. From the
most self-optimization algorithms rely on heuristic control Shannon bound, it can be inferred that the average
rules based on the knowledge of an experienced opera- SINR in a cell (in dB) is a rough approximation of its
tor. Although some advanced self-optimization algorithms average spectral efficiency [34]. Thus, the arithmetic
include unsupervised learning methods (e.g., Q-Learning mean of the average SINR across cells approximates
[16]), simplifying assumptions often cause that there is no the average maximum cell capacity, provided that all
guarantee that optimal system performance is reached. Thus, cells have the same system bandwidth.
the goodness of tuning is usually assessed based on conver-
gence speed and stability issues [31]. Only in very rare cases, (c) Cell harmonic mean [6], where cells with small SINR
the optimality conditions can be reformulated as a control values render the mean value small. Thus, cells with
problem (e.g., [32], where the problem of traffic sharing in worse performance tend to dominate the figure for
GSM is formulated as a balancing problem between adjacent merit.
cells).
In this work, the global CCO problem is formulated as a In this work, the arithmetic mean of the average SINR per
balancing problem by considering the tradeoff between the cell in dB is considered. Thus, the objective function to be
performance of a cell and its neighbors when increasing the maximized is the total system SINR (in dB), computed as
transmit power of the cell under study. From the analysis of
optimality conditions, a new indicator is derived that reflects Maximize Γ𝑡 = ∑Γ (𝑖)
the overall SINR gain in the vicinity of a cell increasing its 𝑖
(1)
transmit power. Thus, it is ensured that changes in power
settings performed on a cell-by-cell basis always improve subject to Γce (𝑖) > Γcemin (𝑖) ,
the overall system performance. The main contributions
of this work are (a) a new indicator used to detect if where Γ(𝑖) and Γce (𝑖) are the average SINR of users in cell 𝑖
increasing the transmit power of a particular base station and in the cell-edge of cell 𝑖, respectively. In (1), Γcemin (𝑖) is the
increases (or decreases) the total system SINR in DL, (b) an minimal value of Γce required in cell 𝑖. The decision variables
algorithm for adjusting base station transmit power in a LTE are cell transmit powers, 𝑃𝑇𝑋 (𝑖), which determine the value
network to increase the coverage area and overall spectral of Γ(𝑖). In this work, 𝑃𝑇𝑋 (𝑖) is defined as the transmit power
efficiency based on the previous indicator, and (c) a thor- level per Physical Resources Block (PRB). It is assumed that
ough comparison of the proposed algorithm with classical 𝑃𝑇𝑋 (𝑖) is the same for all PRBs.
CCO algorithms in a realistic scenario taken from a live In the absence of constraints, any local minimum must
network. satisfy the stationary condition, that is,

𝜕Γ𝑡
3. Problem Formulation = 0 ∀𝑖. (2)
𝜕𝑃𝑇𝑋 (𝑖)
In the following paragraphs, variables in logarithmic units If it is assumed that changing the transmit power of a cell
are written in uppercase, whereas variables in natural units only affects a limited number of neighbor cells, the stationary
are written in lowercase. Likewise, the term DL is omitted condition can be rewritten as
hereafter for brevity.
CCO should be treated as a classical multiobjective 𝜕 (Γ (𝑖) + ∑𝑗∈𝑁(𝑖) Γ (𝑗))
𝜕Γ𝑡
optimization problem, since both coverage and capacity must ≈ , (3)
be maximized. However, it is common practice to assign 𝜕𝑃𝑇𝑋 (𝑖) 𝜕𝑃𝑇𝑋 (𝑖)
a higher priority to either coverage or capacity, since both
features cannot be simultaneously optimized. In most cases, where 𝑁(𝑖) is the set of neighbors of cell 𝑖. The terms in the
optimization is focused on network capacity, provided that numerator reflect the tradeoff between the SINR of a cell and
a minimal network coverage is ensured [8]. In network its neighbors. Specifically, increasing the transmit power of a
planning, network capacity is often evaluated in terms of cell increases the SINR of users served by that cell, that is,
spectral efficiency, which is given by the signal quality in
terms of SINR. Nonetheless, several objective functions have 𝜕Γ (𝑖)
> 0, (4)
been proposed in the literature to measure the overall signal 𝜕𝑃𝑇𝑋 (𝑖)
4 Mobile Information Systems

at the expense of decreasing the SINR of users in neighbor where

cells, that is,
𝜕𝑆 (𝑖)
𝜕 (∑𝑗∈𝑁(𝑖) Γ (𝑗)) 𝐹1 = ,
< 0. (5) 𝜕𝑃𝑇𝑋 (𝑖)
𝜕𝑃𝑇𝑋 (𝑖)
𝜕𝐼 (𝑖)
𝐹2 = ,
To find a local maximum, a simple gradient ascent 𝜕𝑃𝑇𝑋 (𝑖)
method can be used. For this purpose, the slope of the (11)
objective function with respect to each decision variable must 𝜕 (∑𝑗∈𝑁(𝑖) 𝑆 (𝑗))
be obtained. Such an indicator, 𝛽(𝑖), reflecting the gain in the 𝐹3 = ,
𝜕𝑃𝑇𝑋 (𝑖)
total system SINR obtained by increasing the transmit power
of a cell 𝑖 is estimated as follows. 𝜕 (∑𝑗∈𝑁(𝑖) 𝐼 (𝑗))
𝐹4 = .
𝜕𝑃𝑇𝑋 (𝑖)
3.1. Novel Indicator. For convenience, the total SINR around
a cell (in dB) is first defined as
𝐹1 reflects the impact of 𝑃𝑇𝑋 (𝑖) on the desired signal level in
the modified cell (i.e., that received by users in cell 𝑖 from
Γ𝑐 (𝑖) = Γ (𝑖) + ∑ Γ (𝑗)
𝑗∈𝑁(𝑖)
cell 𝑖). 𝐹2 does the same for the total interference received by
(6) users in cell 𝑖 from neighbor cells, 𝐹3 for the desired signal
= 𝑆 (𝑖) − 𝐼 (𝑖) + ∑ (𝑆 (𝑗) − 𝐼 (𝑗)) , received by users in neighbor cells 𝑗, 𝑗 ≠ 𝑖, and 𝐹4 for the
𝑗∈𝑁(𝑖) interference received by users in neighbors 𝑗. All these factors
are calculated as follows.
where 𝑆(𝑖) and 𝐼(𝑖) are the average desired signal and
interference levels (in logarithmic units) received in cell 𝑖, Source Desired Signal Term, 𝐹1 . The signal level received from
respectively. a cell 𝑖 in location (𝑥, 𝑦) is calculated as
In (6), averages correspond to the aggregation of all users
in a cell. In most network planning tools, potential users are 𝑃𝑅𝑋 (𝑖, 𝑥, 𝑦) = 𝑃𝑇𝑋 (𝑖) − PL (𝑖, 𝑥, 𝑦) , (12)
represented by locations. In a real scenario, each location has
a different probability of a user demanding service from it, where PL(𝑖, 𝑥, 𝑦) is the path loss (including antenna gains)
which is given by the spatial traffic distribution. This can between cell 𝑖 and location (𝑥, 𝑦). In particular, the desired
be taken into account by multiplying the values of each signal level in a location is
variable in a location by a weight function representing traffic
probability in that location as 𝑆 (𝑥, 𝑦) = 𝑃𝑅𝑋 (𝑖, 𝑥, 𝑦) , (𝑥, 𝑦) ∈ 𝐴 (𝑖) . (13)

Γ𝑐 (𝑖) = ∑ 𝑝𝑢 (𝑥, 𝑦) (𝑆 (𝑥, 𝑦) − 𝐼 (𝑥, 𝑦)) Note that the spatial traffic distribution, 𝑝𝑢 (𝑥, 𝑦), does not
(𝑥,𝑦)∈𝐴(𝑖)
depend on power settings. Then, from (7), (12), and (13), it is
(7)
deduced that the impact of changing 𝑃𝑇𝑋 (𝑖) on the desired
+ ∑ ∑ 𝑝𝑢 (𝑥, 𝑦) (𝑆 (𝑥, 𝑦) − 𝐼 (𝑥, 𝑦)) ,
signal level received by users in cell 𝑖 is
𝑗∈𝑁(𝑖) (𝑥,𝑦)∈𝐴(𝑗)

where 𝐴(𝑖) is the service area of cell 𝑖 (i.e., locations served by 𝜕 (∑(𝑥,𝑦)∈𝐴(𝑖) 𝑝𝑢 (𝑥, 𝑦) 𝑆 (𝑥, 𝑦))
cell 𝑖), 𝑝𝑢 (𝑥, 𝑦) is the probability that a user generates traffic 𝐹1 =
𝜕𝑃𝑇𝑋 (𝑖)
from location (𝑥, 𝑦), and 𝑆(𝑥, 𝑦) and 𝐼(𝑥, 𝑦) are the received
signal level from the serving cell and total interference level 𝜕𝑃𝑇𝑋 (𝑖) (14)
from neighbor cells for a UE located at (𝑥, 𝑦), respectively. = ∑ 𝑝𝑢 (𝑥, 𝑦) = ∑ 𝑝 (𝑥, 𝑦)
(𝑥,𝑦)∈𝐴(𝑖)
𝜕𝑃𝑇𝑋 (𝑖) (𝑥,𝑦)∈𝐴(𝑖) 𝑢
Obviously,
= 1.
∑ 𝑝𝑢 (𝑥, 𝑦) = 1 ∀𝑖. (8)
(𝑥,𝑦)∈𝐴(𝑖)
Source Interference Term, 𝐹2 . The interference received in
From (3) and (7), it follows that the sensitivity of Γ𝑡 to location (𝑥, 𝑦) in cell 𝑖 from any other cell 𝑗 ≠ 𝑖 is calculated
changes in 𝑃𝑇𝑋 (𝑖) is the same as that of Γ𝑐 , that is, as

𝜕Γ𝑡 (𝑖) 𝜕Γ𝑐 (𝑖)

𝛽 (𝑖) = ≈ . (9) 𝐼 (𝑥, 𝑦) = 10 ⋅ log10 ( ∑ 𝑝𝑟𝑥 (𝑗, 𝑥, 𝑦) 𝑙 (𝑗) + 𝑛0 ) , (15)
𝜕𝑃𝑇𝑋 (𝑖) 𝜕𝑃𝑇𝑋 (𝑖)
𝑗∈𝑁(𝑖)

For simplicity, the partial derivative in (9) is broken down

in four addends as where 𝑝𝑟𝑥 (𝑗, 𝑥, 𝑦) is the power received at (𝑥, 𝑦) from inter-
fering cell 𝑗 (in natural units), 𝑛0 is the thermal noise (in
𝛽 (𝑖) ≈ 𝐹1 − 𝐹2 + 𝐹3 − 𝐹4 , (10) natural units), and 𝑙(𝑗) is the load of cell 𝑗 (dimensionless).
Mobile Information Systems 5

The average load of a cell 𝑖 is estimated by the sum of the traffic From (17), it is deduced that the slope of the spectral efficiency
load generated by each location (𝑥, 𝑦) served by cell 𝑖 as with respect to the transmit power is nonzero only in those
locations where 𝛾min ≤ 𝛾(𝑥, 𝑦) ≤ 𝛾max , so that
𝑙 (𝑖) = ∑ 𝑙 (𝑥, 𝑦) 𝜕𝑙 (𝑗) 1
(𝑥,𝑦)∈𝐴(𝑖) =−
𝜕𝑃𝑇𝑋 (𝑖) 𝑁prb (𝑗)
(16)
∑(𝑥,𝑦)∈𝐴(𝑖) (𝑡 (𝑥, 𝑦) /se (𝑥, 𝑦))
= , 1
𝑁prb (𝑖) ⋅ ∑ (𝑡 (𝑥, 𝑦) 2
(𝑥,𝑦)∈𝐴(𝑗), 𝛼IL [log2 (1 + 𝛾 (𝑥, 𝑦))]
𝛾(𝑥,𝑦)∈[𝛾min ,𝛾max ]
where 𝑡(𝑥, 𝑦) is the average traffic generated by location (𝑥, 𝑦)
Γ(𝑥,𝑦)/10
(in bps), se(𝑥, 𝑦) is the spectral efficiency obtained by users at 1 1 𝜕 [10 ]
location (𝑥, 𝑦) (in bps per PRB), and 𝑁prb (𝑖) is the number of ⋅ )
1 + 𝛾 (𝑥, 𝑦) ln 2 𝜕𝑃𝑇𝑋 (𝑖) (20)
PRBs in cell 𝑖, given by the system bandwidth. In this work,
spectral efficiency is estimated from SINR by the truncated =− ∑ 𝑙 (𝑥, 𝑦)
Shannon bound formula [34] (𝑥,𝑦)∈𝐴(𝑗),
𝛾(𝑥,𝑦)∈[𝛾min ,𝛾max ]

se (𝑥, 𝑦) 𝛾 (𝑥, 𝑦) ln 10
⋅
(1 + 𝛾 (𝑥, 𝑦)) ln (1 + 𝛾 (𝑥, 𝑦)) 10
{
{ 0 𝛾 (𝑥, 𝑦) < 𝛾min ,
{
{ (17)
= {𝛼IL log2 (1 + 𝛾 (𝑥, 𝑦)) 𝛾min ≤ 𝛾 (𝑥, 𝑦) ≤ 𝛾max , 𝜕 (−𝐼 (𝑥, 𝑦))
{ ⋅ .
{
{ 𝜕𝑃𝑇𝑋 (𝑖)
{semax 𝛾max < 𝛾 (𝑥, 𝑦) ,
To compute the derivative of Γ(𝑥, 𝑦) in (20), it has
been considered that only the received interference, 𝐼(𝑥, 𝑦),
where semax is the maximum spectral efficiency that can be changes with 𝑃𝑇𝑋 (𝑖). For tractability, additional side effects
obtained in a location, 𝛾(𝑥, 𝑦) is the SINR in a location (in have not been considered. In addition, it is assumed that the
linear units), 𝛾min and 𝛾max are SINR values corresponding sum of contributions to the change in load from all locations
to 0 and semin , respectively, and 𝛼IL is an attenuation factor (𝑥, 𝑦) in a cell can be approximated by the change in load
representing implementation losses. experienced when all the traffic in the cell is generated in a
Thus, the impact of changing the power of a cell on the single location with the average link performance of the cell.
total interference level received by that cell is This consideration assumes that the change in the average cell
performance is representative of the changes in all points in
𝜕𝐼 (𝑖) 𝜕𝐼 (𝑥, 𝑦) the cell. Thus, (20) can be approximated by
𝐹2 = = ∑ 𝑝𝑢 (𝑥, 𝑦)
𝜕𝑃𝑇𝑋 (𝑖) (𝑥,𝑦)∈𝐴(𝑖) 𝜕𝑃𝑇𝑋 (𝑖) 𝜕𝑙 (𝑗)
𝜕𝑃𝑇𝑋 (𝑖)
[ 10 𝑝𝑢 (𝑥, 𝑦) (21)
= ∑ (18) 𝛾 (𝑗) ln 10 𝜕 (𝐼 (𝑗))
(𝑥,𝑦)∈𝐴(𝑖)
ln 10 ∑𝑗∈𝑁(𝑖) 𝑝𝑟𝑥 (𝑗, 𝑥, 𝑦) 𝑙 (𝑗) + 𝑛0 ≈ 𝑙 (𝑗) ,
[ (1 + 𝛾 (𝑗)) ln (1 + 𝛾 (𝑗)) 10 𝜕𝑃𝑇𝑋 (𝑖)
𝜕𝑙 (𝑗) ] where 𝛾(𝑗) and 𝐼(𝑗) are, respectively, the traffic-weighted
⋅ ∑ 𝑝𝑟𝑥 (𝑗, 𝑥, 𝑦) .
𝑗∈𝑁(𝑖)
𝜕𝑃𝑇𝑋 (𝑖) average SINR (in linear units) and interference level (in
] logarithmic units) for all locations in cell 𝑗. Then,

In (18), it is observed that changing the transmit power of a 𝜕𝐼 (𝑗) 𝜕𝐼 (𝑥, 𝑦)

= ∑ 𝑝 (𝑥, 𝑦)
cell only affects the received interference level in the same cell 𝜕𝑃𝑇𝑋 (𝑖) (𝑥,𝑦)∈𝐴(𝑗) 𝑢 𝜕𝑃𝑇𝑋 (𝑖)
through changes in the load of neighbor cells.
As traffic demand does not depend on link quality, it is
deduced from (16) that the sensitivity of neighbor load to [ 10 𝑝𝑢 (𝑥, 𝑦)
= ∑
changes in 𝑃𝑇𝑋 (𝑖) is (𝑥,𝑦)∈𝐴(𝑗)
ln 10 ∑𝑚∈𝑁(𝑗) 𝑝𝑟𝑥 (𝑚, 𝑥, 𝑦) 𝑙 (𝑚) + 𝑛0 (22)
[

𝜕𝑙 (𝑗) 𝜕 (𝑝𝑟𝑥 (𝑚, 𝑥, 𝑦) 𝑙 (𝑚)) ]

⋅ ∑ .
𝜕𝑃𝑇𝑋 (𝑖) 𝑚∈𝑁(𝑗)
𝜕𝑃𝑇𝑋 (𝑖)
(19) ]
∑(𝑥,𝑦)∈𝐴(𝑗) 𝑡 (𝑥, 𝑦) (𝜕 (1/se (𝑥, 𝑦)) /𝜕𝑃𝑇𝑋 (𝑖)) Note that any change of transmit power in a cell affects signal
= .
𝑁prb (𝑗) quality, spectral efficiency, and, ultimately, the load of the
6 Mobile Information Systems

cell whose power is modified. Such a change in cell load 𝜕𝑝𝑟𝑥 (𝑖, 𝑥, 𝑦)
⋅( 𝑙 (𝑖)
cause changes in the interference generated on neighbor cells, 𝜕𝑃𝑇𝑋 (𝑖)
which also change their signal quality, spectral efficiency, and
load. For tractability, it is assumed here that the only cell
𝜕𝑙 (𝑖)
changing load is the cell changing transmit power. This is true + 𝑝𝑟𝑥 (𝑖, 𝑥, 𝑦) )] .
if the change in transmit power is small enough. It has been 𝜕𝑃𝑇𝑋 (𝑖)
checked with simulations in a typical scenario that neighbors’ (24)
load changes less than 7% (in relative terms) if the change of
transmit power is less than 1 dB. If the change of cell load in The two partial derivatives in (24) show how the interfer-
neighbors is negligible, ence in neighbors changes with the change of transmit power
and load in the modified cell. These terms can be calculated
𝜕𝑙 (𝑚)
≈ 0 ∀𝑚 ≠ 𝑖. (23) as
𝜕𝑃𝑇𝑋 (𝑖)
𝜕𝑝𝑟𝑥 (𝑖, 𝑥, 𝑦) ln 10
Thus, (22) can be rewritten as = 𝑝 (𝑖, 𝑥, 𝑦) , and (25)
𝜕𝑃𝑇𝑋 (𝑖) 10 𝑟𝑥
𝜕𝐼 (𝑗) 10
≈ ∑ [
𝜕𝑃𝑇𝑋 (𝑖) (𝑥,𝑦)∈𝐴(𝑗) ln 10 𝜕𝑙 (𝑖) ln 10 𝛾 (𝑖)
=− 𝑙 (𝑖) . (26)
𝜕𝑃𝑇𝑋 (𝑖) 10 (1 + 𝛾 (𝑖)) ln (1 + 𝛾 (𝑖))
𝑝𝑢 (𝑥, 𝑦)
⋅
∑𝑚∈𝑁(𝑗) 𝑝𝑟𝑥 (𝑚, 𝑥, 𝑦) 𝑙 (𝑚) + 𝑛0 From (21), (24), (25), and (26), it is obtained that

[ 𝑝𝑢 (𝑥, 𝑦)
𝐹2 ≈ ∑
(𝑥,𝑦)∈𝐴(𝑖)
∑𝑗∈𝑁(𝑖) 𝑝𝑟𝑥 (𝑗, 𝑥, 𝑦) 𝑙 (𝑗) + 𝑛0
[
(27)
𝛾 (𝑗) 𝑝𝑢 (𝑥, 𝑦) 𝛾 (𝑖)
⋅ ∑ (𝑝𝑟𝑥 (𝑗, 𝑥, 𝑦) 𝑙 (𝑗) ⋅ ∑ [ ⋅ (𝑝𝑟𝑥 (𝑖, 𝑥, 𝑦) 𝑙 (𝑖) − 𝑝𝑟𝑥 (𝑖, 𝑥, 𝑦) 𝑙 (𝑖) )])] .
𝑗∈𝑁(𝑖)
(1 + 𝛾 (𝑗)) ln (1 + 𝛾 (𝑗)) (𝑥,𝑦)∈𝐴(𝑗)
∑𝑚∈𝑁(𝑗) 𝑝𝑟𝑥 (𝑚, 𝑥, 𝑦) 𝑙 (𝑚) + 𝑛0 (1 + 𝛾 (𝑖)) ln (1 + 𝛾 (𝑖))
]

By grouping terms, 𝐹2 can be rewritten as As neither 𝑝𝑢 (𝑥, 𝑦) nor 𝑃𝑅𝑋 (𝑗, 𝑥, 𝑦) depends on 𝑃𝑇𝑋 (𝑖), 𝐹3 =
0.
𝐹2 ≈ [1 − 𝑓 (𝛾 (𝑖))] ∑ [𝑖𝑐 (𝑗, 𝑖) 𝑓 (𝛾 (𝑗)) 𝑖𝑐 (𝑖, 𝑗)] , (28) Neighbor Interference Term, 𝐹4 . Similarly to 𝐹2 , the change in
𝑗∈𝑁(𝑖) neighbor interference can be expressed as

where 𝜕 (∑𝑗∈𝑁(𝑖) 𝐼 (𝑗))

𝐹4 =
𝜕𝑃𝑇𝑋 (𝑖)
𝛾 (𝑖) (32)
𝑓 (𝛾 (𝑖)) = , (29)
(1 + 𝛾 (𝑖)) ln (1 + 𝛾 (𝑖)) 𝜕𝐼 (𝑥, 𝑦)
= ∑ ∑ 𝑝𝑢 (𝑥, 𝑦) .
𝑗∈𝑁(𝑖) (𝑥,𝑦)∈𝐴(𝑗)
𝜕𝑃𝑇𝑋 (𝑖)
and 𝑖𝑐(𝑗, 𝑖) is the average ratio of interference in cell 𝑖 due to
neighbor 𝑗, defined as Note that 𝑁(𝑗) includes cell 𝑖 as an interferer of cell 𝑗. Then,
the partial derivative in (32) must be separated in two groups
𝑝𝑢 (𝑥, 𝑦) 𝑝𝑟𝑥 (𝑗, 𝑥, 𝑦) 𝑙 (𝑗) of cells: the interfering cell 𝑖 and the rest of interfering cells
𝑖𝑐 (𝑗, 𝑖) = ∑ , (30) 𝑚 ≠ 𝑖. Thus,
∑ 𝑝 (𝑘, 𝑥, 𝑦) 𝑙 (𝑘) + 𝑛0
(𝑥,𝑦)∈𝐴(𝑖) 𝑘∈𝑁(𝑖) 𝑟𝑥

𝜕𝐼 (𝑥, 𝑦) 10 1
Neighbor Desired Signal Term, 𝐹3 . From (7) and (11), =
𝜕𝑃𝑇𝑋 (𝑖) ln 10 ∑𝑚∈𝑁(𝑗) 𝑝𝑟𝑥 (𝑚, 𝑥, 𝑦) 𝑙 (𝑚) + 𝑛0

𝜕 (∑𝑗∈𝑁(𝑖) 𝑆 (𝑗)) 𝜕 (∑𝑚∈𝑁(𝑗),𝑚=𝑖̸ 𝑝𝑟𝑥 (𝑚, 𝑥, 𝑦) 𝑙 (𝑚))

𝐹3 = ⋅[ (33)
𝜕𝑃𝑇𝑋 (𝑖) 𝜕𝑃𝑇𝑋 (𝑖)
(31)
𝜕 (∑𝑗∈𝑁(𝑖) ∑(𝑥,𝑦)∈𝐴(𝑗) 𝑝𝑢 (𝑥, 𝑦) 𝑃𝑅𝑋 (𝑗, 𝑥, 𝑦)) 𝜕 (𝑝𝑟𝑥 (𝑖, 𝑥, 𝑦) 𝑙 (𝑖))
= . + ].
𝜕𝑃𝑇𝑋 (𝑖) 𝜕𝑃𝑇𝑋 (𝑖)
Mobile Information Systems 7

Following the same steps as for 𝐹2 , it is obtained that An output range [𝑃𝑇𝑋min (𝑖), 𝑃𝑇𝑋max (𝑖)] is defined. 𝑃𝑇𝑋max is
usually fixed by operators to the maximum value supported
by the equipment, as network coverage is thus maximized.
𝐹4 ≈ [1 − 𝑓 (𝛾 (𝑖))] ∑ [𝑖𝑐 (𝑖, 𝑗) Such a maximum value may be hardware (i.e., power ampli-
𝑗∈𝑁(𝑖)
[ fier) or software (i.e., license) limited. Note that 𝑃𝑇𝑋max is not
(34) necessarily the same in all cells, since the DL power amplifier
and/or the licensed power may be different between cells. In
+ ∑ 𝑖𝑐 (𝑚, 𝑗) 𝑓 (𝛾 (𝑚)) 𝑖𝑐 (𝑖, 𝑚)] . contrast, 𝑃𝑇𝑋min is defined as a safety brake for the tuning
𝑚∈𝑁(𝑗),𝑚=𝑖̸
] process to prevent coverage issues. In this work, 𝑃𝑇𝑋min is set to
10 dB less than 𝑃𝑇𝑋max for each cell to avoid excessive transmit
Two terms can be identified in (34). The first term reflects the power reduction.
change of interference in cell 𝑗, caused by the load variation
in cell 𝑖 due to the change in spectral efficiency from the
new transmit power of cell 𝑖. The second term reflects the
5. Performance Assessment
change of the interference received by cell 𝑗, caused by the The self-planning algorithm is validated with a static system-
load variation in the rest of interferers of cell 𝑗, which also level LTE simulator implementing a live scenario adjusted
modify their load due to change of interference from cell 𝑖. with real measurements. The assessment methodology is first
described and results are presented later.
4. Self-Planning Algorithm
5.1. Assessment Methodology. This section describes the sim-
In this section, a heuristic method to find the best change ulation scenario and the experiments carried out to assess the
in DL transmit power on a cell basis is presented. The proposed indicator and self-planning method.
aim of the method is to adjust the transmit power of each
cell in the system so as to improve the total system SINR,
Γ𝑡 . The method makes use of a classical gradient ascent 5.1.1. Simulation Scenario. A DL static system-level LTE
algorithm to find the best tradeoff between the SINR of a cell simulator implementing a real macrocellular scenario has
and its neighbors, based on the above-described indicator, been developed in MATLAB [30]. The simulator is designed
𝛽(𝑖), reflecting whether increasing the transmit power of to make the most of available network statistics to model a
a particular base station increases (or decreases) the total live macrocellular scenario. For this purpose, the simulator
system SINR. includes the following features:
The gradient ascent algorithm is designed as a set of sim- (a) Initialization of cell load distribution across the sce-
ple proportional controllers (one per cell), which iteratively nario with PRB utilization figures obtained from
compute changes in the DL transmit power based on the counters in the network management system of a live
value of the 𝛽 indicator. Specifically, the output of one of LTE network.
these controllers is the change in DL transmit power (in dB),
Δ𝑃𝑇𝑋 (𝑖), computed as (b) Adjustment of spatial user distribution within a cell by
distance rings, so that the probability of a user being in
a distance ring is derived from Timing Advance (TA)
{
{ −1 𝛽 (𝑖) ≤ 𝛽min ,
{
{ distributions [35].
Δ𝑃𝑇𝑋 (𝑖) = {0 𝛽min < 𝛽 (𝑖) < 𝛽max , (35) (c) Tuning of propagation model parameters according
{
{
{ to live RSRP measurements statistics.
{1 𝛽 (𝑖) ≥ 𝛽max ,
Table 1 summarizes the main scenario parameters. The
where 𝛽min and 𝛽max are thresholds for triggering 𝑃𝑇𝑋 (𝑖) simulated area comprises 129 cells distributed in 44 sites
modifications, introduced to keep the number of network with an average Inter-Site Distance (ISD) of 0.8 km. This
changes to a minimum. In this work, both thresholds are scenario, with a relatively low ISD, is representative of an
symmetrical; that is, 𝛽max = −𝛽min . interference-limited scenario in a dense urban area. The
The proposed algorithm works as an iterative process, geographical area under analysis is divided into a regular
starting from an initial power value, 𝑃𝑇𝑋 (𝑖), which is later grid of points, representing potential user locations. Received
modified at every iteration (referred to as optimization loop). signal level at each point from base stations is computed by a
Every loop starts with the collection of network statistics, macrocellular propagation model including log-normal slow
from which 𝛽 is computed for all cells in the scenario. The fading (i.e., Winner II C2 model [33]). For this purpose, user
collection period must be large enough (e.g., one day) to locations are classified into Line Of Sight (LOS) or Non-
improve the robustness of the algorithm. Then, (35) is used Line of Sight (NLOS) conditions based on real geolocated
(𝑙)
to compute power changes, Δ𝑃𝑇𝑋 (𝑖), where 𝑙 is the loop data of buildings and antenna positions in the scenario. Grid
number. Finally, the new values of 𝑃𝑇𝑋 (𝑖) for the next loop resolution is 40 meters. Cell service areas are computed by
are calculated as combining path losses and antenna gains, so that the serving
cell is that providing the maximum pilot signal level for each
(𝑙+1) (𝑙) (𝑙)
𝑃𝑇𝑋 (𝑖) = 𝑃𝑇𝑋 (𝑖) + Δ𝑃𝑇𝑋 (𝑖) . (36) point. Likewise, neighbor cells are defined as those providing
8 Mobile Information Systems

Table 1: Simulation scenario.

Number of sites 44
Number of cells 129
Avg. intersite distance [m] 815
Carrier frequency [MHz] 734
System bandwidth [MHz] 10
Number of PRBs 50
UE height [m] 1.5
BS antenna height [m] [3, 54]
Initial cell transmit power [dBm] [46.5, 47.4]
Maximum antenna gain [dB] 15
Antenna tilt angle [ ∘ ] [0, 13]
Propagation model Winner II C2 [33] with 𝑋 = 14 for NLOS users
Grid resolution [m] 40
Spatial traffic distribution Distance based on TA measurements
PRB utilization ratio [%] [5, 70]

the largest signal levels in the service area of a cell. Then, in the simulator are not recalculated when power settings are
interference level at every point is computed by adding the modified.
interference from all neighbor cells. Finally, 𝛽 is calculated
on a cell basis and the self-planning algorithm is executed
to obtain new 𝑃𝑇𝑋 values for the next optimization loop. 5.1.3. Algorithm Assessment. Three power planning methods
PRB utilization ratios are updated at every new loop by are compared. A first method is the DL transmit power plan
estimating the impact of 𝑃𝑇𝑋 (𝑖) changes on the radio link originally implemented by the operator, denoted as operator
spectral efficiency at every cell. The reader is referred to [30] solution (OS). OS is the baseline against which all other
for a more detailed explanation of the simulation tool. methods are compared. In OS plan, DL transmit power is
set to the maximum value in all cells. This is due to the
fact that DL transmit power is often configured when the
5.1.2. Validation of 𝛽 Indicator. A sensitivity test is first carried site is launched and remains unchanged when new sites are
out to check the accuracy of indicator 𝛽, reflecting the impact deployed in the surroundings, which may generate useless
of changing the power of a cell on the total system SINR. cell overlapping and high intercell interference.
For this purpose, the default power plan configured by the A second method is the iterative self-planning algorithm
operator is modified by increasing the power of a single cell by for DL transmit power proposed here, denoted as SINR-
1 dB, so a new power plan is obtained. This process is repeated PWR. This method is initialized with the OS plan and 30
for every cell in the scenario, so 129 new power plans are optimization loops are simulated. It is checked a posteriori
constructed. Then, the difference in the total system SINR, that this number of loops is enough to reach equilibrium.
ΔΓ𝑡 (𝑖), from each individual power change is computed by A third method is the iterative self-planning algorithm for
subtracting the values with the default and the new power remote electrical tilt (RET) based on trace files described in
plan. Such differences should coincide with the values of 𝛽 for [30], denoted as TF-RET (for Trace-based Fuzzy). TF-RET
the different cells, 𝛽(𝑖), estimated from network performance adjusts antenna tilt values instead of base station transmit
data obtained with the default plan as in (9). power. For this purpose, three performance indicators are
It should be pointed out that the simulation tool can obtained from connection traces to detect cell overshooting,
accurately model cell coupling effects. For instance, SINR in useless cell overlapping, and cell-edge coverage problems.
neighbor 𝑗 is modified when the power of a cell 𝑖 is modified. Then, downtilting is performed in those cells generating over-
As a consequence, spectral efficiency in cell 𝑗 changes, and so shooting and/or useless overlapping, and without coverage
does its PRB utilization ratio. This load change in cell 𝑗 causes problems. As in [30], this method is also initialized with the
interference changes in both cell 𝑖 and other neighboring OS plan and 20 optimization loops are simulated. It is checked
cells. Thus, the initial power change in cell 𝑖 is propagated a posteriori that this number of loops is enough to reach
across the network. Likewise, the simulator updates cell equilibrium.
service areas with the new power setting. However, these For brevity, the analysis of SINR-PWR and TF-RET is
side effects cannot be taken into account by the analytical restricted to the solution obtained in the last iteration.
approach used to derive 𝛽. On the contrary, the definition of 𝛽 Two figures of merit are used to assess power planning
only considers first-order effects of changing power settings. algorithms:
For a fair comparison, and only for this experiment, PRB
utilization of neighbor cells 𝑗 that do not change power is
kept unaltered in the simulations, that is, only the modified (i) Overall average DL SINR, SINRavg , as a measure of
cell 𝑖 changes its PRB utilization. Likewise, cell service areas network connection quality and spectral efficiency,
Mobile Information Systems 9

1.5 14.2

14
1
13.8
ΔΓt (dB) (simulated)

0.5 13.6

３）．２；ＰＡ (dB)
y = 0.93x + 0.05 13.4
0
R2 = 0.98 13.2
−0.5 13

12.8
−1
12.6
−1.5
−1.5 −1 −0.5 0 0.5 1 1.5 12.4
1.6 1.8 2 2.2 2.4 2.6 2.8 3
 (dB) (analytical)
３）．２＝？ (dB)
Figure 1: Analytical and simulated change of total DL SINR.
OS
SINR−PWR
TF−RET
calculated as the arithmetic mean of the average DL
SINR in each cell, Γ(𝑖), and Figure 2: Performance comparison.
(ii) Overall cell-edge DL SINR, SINRce , as a measure of
network coverage, calculated as the arithmetic mean
the same transmit power change leads to a decrease in the
of the 5th-percentile DL SINR in each cell, Γce (𝑖).
total system SINR, denoting a tightly coupled cell scenario.
In practice, any change in power settings through opti- This result justifies the need for adjusting transmit power on
mization loops might cause that some locations in the a cell-by-cell basis.
scenario do not receive enough signal level for establishing
a connection, or, conversely, some locations initially not 5.2.2. Algorithm Assessment. Figure 2 compares the results
served by any cell could reach enough signal level to start of the different algorithms. The overall cell-edge and cell-
a connection in a later optimization loop. These changes in average SINR metrics are shown on the 𝑥- and 𝑦-axes,
the coverage area would have an influence on the total traffic respectively. Points in the upper right part of the figure
carried by the network. For a fair comparison, both SINR show better performance. The OS configuration method
indicators are computed in the same set of locations along used as a reference is represented by a single dot (dia-
iterations (i.e., the evaluated geographical area is always the mond). Self-planning methods (i.e., SINR-PWR and TF-
same). RET) are represented by a curve of multiple dots showing
the performance of intermediate network parameters set-
5.2. Results. The validation of the 𝛽 indicator is first presented tings reached across optimization loops. For clarity, the last
and the assessment of power planning algorithms is discussed value in these iterative methods is highlighted with a filled
later. marker. Both SINR-PWR and TF-RET curves start with the
OS network configuration, and their performance is thus
5.2.1. Validation of 𝛽 Indicator. Figure 1 shows the accuracy the same in the first iteration. Thereafter, SINR-PWR and
of the proposed indicator by comparing estimates obtained TF-RET improve both average and cell-edge SINR along
with the formulas against results obtained with the simulator. iterations.
Each point in the figure corresponds to one of the new Table 2 compares the performance of the iterative meth-
129 power plans built by increasing the transmit power of ods, SINR-PWR and TF-RET, at the end of the optimiza-
a cell by 1 dB. The 𝑥- and 𝑦-axes represent the values of 𝛽 tion process against the initial solution, OS. Table 2 also
and ΔΓ𝑡 , respectively. It is observed that both indicators are presents the overall average and cell-edge DL user through-
strongly correlated, since the coefficient of determination, 𝑅2 , put, UeTHavg and UeTHce , respectively, calculated as the
is 0.98, and the value of the regression slope is close to 1. This arithmetic mean of the individual cell values (average and
similarity between simulated and analytical values proves the cell-edge DL user throughput, resp.). DL user throughput
validity of the 𝛽 indicator. in each user location is computed from SINR values by
Moreover, it is observed that the impact of increasing the the bounded Shannon formula [34] with attenuation factor,
transmit power of a cell greatly varies from cell to cell. In 𝛼imp = 0.6, assuming that the whole system bandwidth is
some cells (up and right points in Figure 1), the 1-dB transmit available to the user (i.e., 50 PRB ⋅ 180 kHz/PRB = 9 MHz). In
power increase is directly translated into a 1-dB increase in addition, other important indicators are also included, such
the total system SINR denoting an isolated cells scenario. In as the average received pilot signal level from the serving cell,
contrast, in other cells (left and down points in the figure), RSRP, average DL interference level, 𝐼, average deviation of
10 Mobile Information Systems

Table 2: Method performance at the end of the adjusting process.

OS SINR-PWR TF-RET
SINRavg [dB] 12.69 13.90 13.70
SINRce [dB] 1.72 2.91 2.86
UeTHavg [Mbps] 23.36 25.36 25.02
UeTHce [Mbps] 7.44 8.74 8.70
RSRP [dBm] −88.44 −89.90 −88.17
𝐼 [dBm] −101.13 −103.80 −101.88
Δ𝑃𝑇𝑋 [dB] — −1.80 —
Δ𝛼[ ∘ ] — — 0.50
Number of modified cells — 48 65

transmit power and tilt angle from initial settings, Δ𝑃𝑇𝑋 and performance indicators, and only 0.085 seconds per loop is
Δ𝛼, and the number of modified cells. spent on the computation of the proposed indicator (i.e., less
In Table 2, it is observed that both methods outper- than 0.13% of the loop execution time).
form the current operator solution. Specifically, SINR-PWR Moreover, by computing the 𝛽 indicator analytically,
improves SINRavg by 1.21 dB and SINRce by 1.19 dB when instead of deriving it by perturbation analysis, the number
compared to OS, whereas TF-RET improves those indicators of parameter plans to be simulated is reduced by 𝑁𝑐 times.
by only 1.01 and 1.14 dB, respectively. It is also observed that Note that the 𝛽 indicator can be calculated analytically by
SINR-PWR outperforms TF-RET, since a better performance simulating a single parameter plan, whereas a perturbation
is achieved for both indicators at the end of the optimization analysis needs changing the power setting of one cell at a time
process (i.e., 13.90 against 13.70 dB for SINRavg , and 2.91 and simulating a new parameter plan per cell.
against 2.86 dB for SINRce ). SINR improvements are directly
translated into a better user experience reflected in through- 6. Conclusions
put indicators. Specifically, UeTHavg and UeTHce indicators
are increased by 8.6% and 17.5% in relative terms, respectively, In this work, a novel self-planning algorithm has been
with SINR-PWR algorithm. In contrast, TF-RET algorithm presented for adjusting the transmit power of LTE base
only achieves a 7.1% and 16.9% increase for those indicators. stations on a cell basis to improve both network coverage and
The reason for the better performance of SINR-PWR is the overall spectral efficiency. The algorithm is designed as a set
fact that SINR-PWR is based on an analytical approach that of independent controllers that decide whether to increase
ensures a (local) maximum of system performance, whereas or decrease the transmit power of a cell based on a new
TF-RET is based on heuristic rules. Recall that the main indicator showing the expected impact of that change on the
goal of SINR-PWR is to improve the overall average SINR, total system DL SINR. The proposed self-planning algorithm
regardless of the overall cell-edge SINR. Then, the fact that has been tested in a static system-level simulator modeling
SINR-PWR also improves SINRce , and more than TF-RET, a live dense urban interference-limited LTE scenario. Results
is a positive side effect that proves the robustness of the have shown that the proposed method can improve both cell-
proposed method. average and cell-edge SINR values by more than 1 dB when
A close inspection of Table 2 shows that SINR-PWR compared with the solution currently implemented in the
network.
obtains its results by decreasing both RSRP and 𝐼 values
During the tests, the proposed method has been com-
(−1.46 and −2.67 dB, resp.). Such a reduction, typical of
pared with a state-of-the-art self-planning method based on
an interference-limited scenario, is done without deteriorat-
adjusting antenna tilts. A priori, tilting is a more powerful
ing the overall SINR cell-edge performance, which is also
technique, as it can improve both the modified cell (with
improved. In contrast, TF-RET maintains RSRP and 𝐼 values. higher desired signal level) and its neighbors (with less
As an additional advantage, SINR-PWR modifies fewer cells interference). However, the proposed power-based method
than TF-RET (i.e., 48 power changes versus 65 up/downtilts). outperforms the tilt-based method. The reason for that supe-
riority is the fact that the power-based method is designed
5.3. Implementation Issues. The time complexity of SINR- to reach the optimal power plan, based on the analysis of
PWR algorithm is O(𝑁𝑐 ), where 𝑁𝑐 is the number of cells optimality conditions. Thus, the proposed approach ensures
in the analyzed area. The method is designed as a control that network performance is always improved after every
algorithm and therefore has a low computational complexity. change of power settings. In contrast, the tilt-based method
Specifically, the total execution time of 30 optimization loops is based on a heuristic approach, in the absence of an
in the considered scenario with 129 cells, in a computer analytic expression of the optimality conditions of antenna
with a clock frequency of 3.47 GHz and 12 GB of RAM, is tilting. Moreover, unlike antenna tilting, power adjustment
1980 seconds (66 seconds per loop on average). Most of this is still valid for cells with omnidirectional or multiband
time is spent on simulating the scenario to obtain network antennas.
Mobile Information Systems 11

Unlike other self-planning approaches, the proposed [10] A. B. Vallejo-Mora, M. Toril, S. Luna-Ramirez, A. Mendo,
method is designed as a control algorithm. Due to its low and S. Pedraza, “Congestion relief in subway areas by tuning
computational complexity, the method can be adapted to be uplink power control in LTE,” IEEE Transactions on Vehicular
used as a self-optimization algorithm, provided that the 𝛽 Technology, vol. 66, no. 7, pp. 6489–6497, 2017.
indicators are computed from network performance mea- [11] D. Kim, “Downlink power allocation and adjustment for
surements available in connection traces. For this purpose, CDMA cellular systems,” IEEE Communications Letters, vol. 1,
periodic RSRP measurements should be activated in all sites. no. 4, pp. 96–98, 1997.
This data should be processed periodically in a centralized [12] D. Kim, “A simple algorithm for adjusting cell-site transmitter
node to obtain the new transmit power settings by the power in CDMA cellular systems,” IEEE Transactions on Vehic-
algorithm. Alternatively, each base station might exchange ular Technology, vol. 48, no. 4, pp. 1092–1098, 1999.
performance measurements with its neighbors to derive the [13] T. Cai, G. P. Koudouridis, C. Qvarfordt, J. Johansson, and
associated 𝛽 indicator in a distributed fashion. P. Legg, “Coverage and capacity optimization in E-UTRAN
based on central coordination and distributed Gibbs sampling,”
in Proceedings of the 2010 IEEE 71st Vehicular Technology
Conflicts of Interest Conference, VTC 2010-Spring, Taipei, Taiwan, May 2010.
[14] K. Majewski and M. Koonert, “Analytic uplink cell load approx-
The authors declare that they have no conflicts of interest. imation for planning fractional power control in LTE networks,”
Telecommunication Systems, vol. 52, no. 2, pp. 1081–1090, 2013.
Acknowledgments [15] A. Engels, M. Reyer, X. Xu, R. Mathar, J. Zhang, and H. Zhuang,
“Autonomous self-optimization of coverage and capacity in LTE
This work has been funded by the Spanish Ministry of Econ- cellular networks,” IEEE Transactions on Vehicular Technology,
omy and Competitiveness (TEC2015-69982-R) and Optimi- vol. 62, no. 5, pp. 1989–2004, 2013.
Ericsson. [16] S. Fan, H. Tian, and C. Sengul, “Self-optimization of coverage
and capacity based on a fuzzy neural network with cooperative
reinforcement learning,” Eurasip Journal on Wireless Communi-
References cations and Networking, vol. 2014, article no. 57, 2014.
[17] J. Ramiro and K. Hamied, Self-Organizing Networks: Self-
[1] 3GPP TS 32.500, “Telecommunication management; Self-
Planning, Self-Optimization and Self-Healing for GSM, UMTS
Organizing Networks (SON); Concepts and requirements,”
and LTE, Wiley, Hoboken, NJ, USA, 2011.
2009, Version 9.0.0.
[18] J. Lempiäinen and M. Manninen, UMTS Radio Network Plan-
[2] NGMN, “Next Generation Mobile Networks Recommendation
ning, Optimization and QOS Management, Kluwer Academic
on SON and O&M Requirements,” Req. Spec. v1, vol. 23, 2008.
Publishers, Dordrecht, Netherlands, 2003.
[3] S. Sesia, I. Toufik, and M. Baker, LTE, The UMTS Long Term [19] I. Luketic, D. Simunic, and T. Blajic, “Optimization of coverage
Evolution: From Theory to Practice, Wiley, Hoboken, NJ, USA, and capacity of Self-Organizing Network in LTE,” in Proceedings
2009. of the 34th International Convention MIPRO, pp. 612–617,
[4] J. Whitehead, “Signal-level-based dynamic power control for Opatija, Croatia, 2011.
co-channel interference management,” in Proceedings of the [20] J. Zhang, C. Sun, Y. Yi, and H. Zhuang, “A hybrid frame-
IEEE 43rd Vehicular Technology Conference, pp. 499–502, work for capacity and coverage optimization in self-organizing
Secaucus, NJ, USA, 1993. LTE networks,” in Proceedings of the 2013 IEEE 24th Annual
[5] A. Simonsson and A. Furuskär, “Uplink power control in International Symposium on Personal, Indoor, and Mobile Radio
LTE—overview and performance: principles and benefits of Communications, PIMRC 2013, pp. 2919–2923, London, UK,
utilizing rather than compensating for SINR variations,” in September 2013.
Proceedings of the 68th Semi-Annual IEEE Vehicular Technology [21] A. Saeed, O. G. Aliu, and M. A. Imran, “Controlling self healing
(VTC ’08), pp. 1–5, Calgary, Canada, September 2008. cellular networks using fuzzy logic,” in Proceedings of the 2012
[6] A. Awada, B. Wegmann, I. Viering, and A. Klein, “Optimizing IEEE Wireless Communications and Networking Conference,
the radio network parameters of the long term evolution WCNC 2012, pp. 3080–3084, Shanghai, China, April 2012.
system using Taguchi’s method,” IEEE Transactions on Vehicular [22] J. Moysen and L. Giupponi, “A reinforcement learning based
Technology, vol. 60, no. 8, pp. 3825–3839, 2011. solution for self-healing in LTE networks,” in Proceedings of the
[7] M. Dirani and Z. Altman, “Self-organizing networks in next 80th IEEE Vehicular Technology Conference (VTC Fall ’14), pp.
generation radio access networks: application to fractional 1–6, IEEE, Vancouver, Canada, September 2014.
power control,” Computer Networks, vol. 55, no. 2, pp. 431–438, [23] O. Onireti, A. Zoha, J. Moysen et al., “A cell outage management
2011. framework for dense heterogeneous networks,” IEEE Transac-
[8] J. Á. Fernández-Segovia, S. Luna-Ramı́rez, M. Toril, A. B. tions on Vehicular Technology, vol. 65, no. 4, pp. 2097–2113, 2016.
Vallejo-Mora, and C. Úbeda, “A computationally efficient [24] U. Turke and M. Koonert, “Advanced site configuration tech-
method for self-planning uplink power control parameters niques for automatic UMTS radio network design,” in Proceed-
in LTE,” EURASIP Journal on Wireless Communications and ings of the 2005 IEEE 61st Vehicular Technology Conference, pp.
Networking, vol. 2015, no. 1, article 80, 2015. 1960–1964, Stockholm, Sweden, 2005.
[9] J. A. Fernández-Segovia, S. Luna-Ramı́rez, M. Toril, and C. [25] J. Niemelä, T. Isotalo, and J. Lempiäinen, “Optimum antenna
Úbeda, “A fast self-planning approach for fractional uplink downtilt angles for macrocellular WCDMA network,” Eurasip
power control parameters in LTE networks,” Mobile Information Journal on Wireless Communications and Networking, vol. 2005,
Systems, vol. 2016, Article ID 8267407, 11 pages, 2016. no. 5, pp. 816–827, 2005.
12 Mobile Information Systems

[26] I. Siomina, P. Värbrand, and Y. Di, “Automated optimization

of service coverage and base station antenna configuration in
UMTS networks,” IEEE Wireless Communications, vol. 13, no. 6,
pp. 16–25, 2006.
[27] W. Jianhui and Y. Dongfeng, “Antenna downtilt performance
in urban environments,” in Proceedings of the 1996 15th Annual
Military Communications Conference (MILCOM ’96), vol. 3, pp.
739–744, McLean, Va, USA, 1996.
[28] O. N. C. Yilmaz, J. Hämäläinen, and S. Hämäläinen, “Self-
optimization of remote electrical tilt,” in Proceedings of the
2010 IEEE 21st International Symposium on Personal Indoor
and Mobile Radio Communications, PIMRC 2010, pp. 1128–1132,
Istanbul, Turkey, September 2010.
[29] A. J. Fehske, H. Klessig, J. Voigt, and G. P. Fettweis, “Concurrent
load-aware adjustment of user association and antenna tilts in
self-organizing radio networks,” IEEE Transactions on Vehicular
Technology, vol. 62, no. 5, pp. 1974–1988, 2013.
[30] V. Buenestado, M. Toril, S. Luna-Ramı́rez, J. M. Ruiz-Avilés,
and A. Mendo, “Self-tuning of remote electrical tilts based on
call traces for coverage and capacity optimization in LTE,” IEEE
Transactions on Vehicular Technology, vol. 66, no. 5, pp. 4315–
4326, 2017.
[31] M. Toril and V. Wille, “Optimization of handover parameters for
traffic sharing in GERAN,” Wireless Personal Communications,
vol. 47, no. 3, pp. 315–336, 2008.
[32] S. Luna-Ramı́rez, M. Toril, M. Fernández-Navarro, and V.
Wille, “Optimal traffic sharing in GERAN,” Wireless Personal
Communications, vol. 57, no. 4, pp. 553–574, 2011.
[33] P. Kyösti, J. Meinilä, and L. Hentilä, “IST-WINNER D1.1.2
WINNER II channel models,” 2007, https://www.researchgate
.net/publication/259900906 IST-4-027756 WINNER II D112
v12 WINNER II channel models.
[34] 3GPP, “LTE; Evolved Universal Terrestrial Radio Access (E-
UTRA); Radio Frequency (RF) System Scenarios,” TR 36.942,
version 8.2.0, 2009.
[35] M. Miernik, “Application of 2G spatial traffic analysis in the
process of 2G and 3G radio network optimization,” in Proceed-
ings of the 2007 IEEE 65th Vehicular Technology Conference -
VTC2007-Spring, pp. 909–913, Dublin, Ireland, April 2007.
 Advances in Journal of
Industrial Engineering
Multimedia
Applied
Computational
Intelligence and Soft
Computing
The Scientific International Journal of
Distributed
Hindawi Publishing Corporation
World Journal
Hindawi Publishing Corporation
Sensor Networks
Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 201

Advances in

Fuzzy
Systems
Modelling &
Simulation
in Engineering
Hindawi Publishing Corporation
Hindawi Publishing Corporation Volume 2014 http://www.hindawi.com Volume 2014
http://www.hindawi.com

Submit your manuscripts at

-RXUQDORI
https://www.hindawi.com
&RPSXWHU1HWZRUNV
DQG&RPPXQLFDWLRQV  Advances in 
Artificial
Intelligence
+LQGDZL3XEOLVKLQJ&RUSRUDWLRQ
KWWSZZZKLQGDZLFRP 9ROXPH Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014

International Journal of Advances in

Biomedical Imaging $UWLÀFLDO
1HXUDO6\VWHPV

International Journal of
Advances in Computer Games Advances in
Computer Engineering Technology Software Engineering
Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 201 http://www.hindawi.com Volume 2014

International Journal of
Reconfigurable
Computing

Advances in Computational Journal of

Journal of Human-Computer Intelligence and Electrical and Computer
Robotics
Hindawi Publishing Corporation
Interaction
Hindawi Publishing Corporation
Neuroscience
Hindawi Publishing Corporation Hindawi Publishing Corporation
Engineering
Hindawi Publishing Corporation
http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014 http://www.hindawi.com Volume 2014