Content uploaded by Norsheila Fisal
Author content
All content in this area was uploaded by Norsheila Fisal
Content may be subject to copyright.
Challenges and Practical Implementation of Self-
Organizing Networks in LTE/LTE-Advanced
Systems
M. M. S. Marwangi, N. Fisal, S.K.S. Yusof, Rozeha A. Rashid, Aimi S. A. Ghafar, Faiz A. Saparudin, and N. Katiran
Centre of Excellence (CoE) in Telecommunication Technology
Faculty of Electrical Engineering, Universiti Teknologi Malaysia (UTM)
Skudai, Johor, Malaysia
Email: marwangi.maharum@fkegraduate.utm.my, {sheila, kamilah, rozeha}@fke.utm.my, aimisyamimi@fkegraduate.utm.my,
fasraf2@live.utm.my, norshida@uthm.edu.my
Abstract—Self-organizing network/s (SON) was promoted by the
Third Generation Partnership Project (3GPP) and Next
Generation Mobile Networks (NGMN). SON has been introduced
since in 3GPP Release-8, Release-9 and currently included in
Release-10 framework as an excellent solution that promises
improvements and market potential for future wireless networks.
The purpose of this paper is to provide an overview on SON in
advanced wireless networks such as Long Term Evolution (LTE)
and LTE-Advanced systems. Although SON promises several
advantages for both network operators and users, there are still
challenges in implementing SON network in reality. Therefore,
this paper addresses those possible challenges, discusses about
SON implementation in inter-cell interference of cross-tier
networks and provides future trends of the research area.
Keywords-Rel-8; Rel-9; LTE-Advanced; Self-Organizing
Network/s (SON); challenges; inter-cell interference.
I. INTRODUCTION
Self-organization is a system that consists of both structural
and functional aspects. All the entities involved within this
system are well arranged in particular manner and they interact
among themselves to fulfill the sole purpose of the overall
system [1]. A system which is self-organized may not have any
external or central control entity, but the controlling
mechanisms are distributed and localized among the entities
within the system. Some examples of bio-inspired organization
structure are; ants tend to find the shortest path to the food
sources, and birds will organize themselves in structured
swarms to fly back to their nest in the evening.
In every deployed wireless networks, optimization
activities are certainly needed and must be done in a
continuous and constant manner. That is the reason why most
of the wireless network operators use human workforce for the
operations, administrations, and maintenance (OAM) activities;
such as planning the network, analyzing reports, identifying
problems, improving network mechanisms, and optimizing
decisions. As a result, OAM activities contribute several
percentage to the operational expenditures. To overcome
problems related to capital and operational expenditures, the
network operators are now looking towards having a self-
organized network system. This is where the so called network
(SON) comes up as it is equipped with the capability to
configure, optimize, and heal itself.
Complexity, number of nodes, and homogeneity are the
main reasons why a network should be self-organized [2–4]. It
is because the operating parameters of the base stations become
more complicated and change rapidly as the networks grow
from 2G to 3G and now entering the 4G networks. Hence,
advanced technology needs extra effort to be included inside
the optimization processes or else, the operational expenses
will keep on rising due to the network complexity. Moreover,
recent arrival of small base stations such as femtocells and
picocells in heterogeneous network (Het-Net) deployment had
led to the growing number of nodes. For this reason, sufficient
self-organized effort is needed to manage the increasing
number of nodes with less human involvement since the
traditional OAM approach would not be able to handle this
rising volume. Last but not least, most of 2G, 3G, and 4G
networks might be operated in parallel within some time. In
this scenario, using separate OAM system is not a good choice
while utilizing new OAM system would affect system stability
and add extra costs. Again, all of these factors play important
role in moving towards the necessity of having a network that
is self-organized.
Therefore, SON is extensively discussed in 3GPP
standardization for both LTE and LTE-Advanced systems.
SON was introduced in LTE systems as early as in the first
release (Rel-8) where it covers several aspects of self-
configuration use cases such as automatic inventory, automatic
software download, automatic neighbor relations, and
automatic physical cell identity (PCI) assignment. Additional
use cases such as coverage and capacity optimization, load
balancing optimization, RACH optimization, and mobility
robustness optimization were later discussed in second release
of LTE (Rel-9). Besides that, energy saving and interference
control were also included in Rel-9 framework. The details on
SON use cases description can be found in [5]. The use cases
are also discussed in the following LTE standard releases, for
examples Rel-10 (LTE-Advanced), Rel-11, and so forth. As
illustrated in Figure 1, there are three elements that form SON.
The elements are self-configuration, self-optimization, and
self-healing.
Figure 1. Elements that form SON
First, self-configuration element covers pre-operational
state and is triggered when an intentional event occur. For
example, the self-configured mechanism will automatically
configure the installation process of newly deployed entities in
order to achieve basic configuration for the system operation.
This automated mechanism also allows new evolved NodeB
(eNB) to be integrated via plug-and-play method. On the other
hand, self-optimization element is required during operational
state with the purpose of maximizing network performance and
reducing operational expenditures. Soon after eNB has
completed its pre-operational state, it starts to optimize the
operating procedures such as optimizing the neighbour list and
energy usage of the base station. Finally, self-healing
mechanism allows detected failures during the network
communication for possible correction and healing process. For
instance, as discussed in [4], self-healing of SON Rel-10 will
perform correction on the affected network elements based on
analyzed fault management, alarms, notifications, and self-test
results. Obviously, on-site operations and settings can be
eliminated by having the self-organizing functions.
The subsequent parts of the paper are organized as follows.
Section II highlights the challenges in implementing SON.
Next, section III discusses the implementation of SON in
practical deployment aspect. Then, section IV provides SON as
a future trend in research and finally, concluding remarks
terminate this paper.
II. CHALLENGES IN IMPLEMENTING SON
Obviously, SON offers several advantages. First, it is a
method to improve OAM process by reducing installation and
management costs. The costs can be reduced by applying
automated mechanisms of self-configuration, self-optimization,
and self-healing in the operational tasks [6]. Therefore,
manpower usage for manual configuration processes can be
reduced. Secondly, SON allows for simplification and
optimization of network tasks [2]. According to [7], SON
enables optimization of network elements and physical
resources besides extending the equipment lifetime. Once
again, capital expenditure (CAPEX) is reduced.
Moreover, self-organized based networks have the ability to
configure and optimize automatically where the functions are
in the core network. Hence, human error during OAM manual
monitoring is reduced since their involvement in the
management process is lower. Finally, users may experience
improved signal qualities as the network parameters are well
optimized when network interference and quality degradations
are mitigated from inaccurate planning or equipment failures
[8].
Even though SON is highly in demand, there are still
several challenges in implementing this network in reality. The
challenges were also discussed in [6] and [7].
A. Reliability
System that operates under multi-vendor environment can
be further improved through self-optimization functionality.
This can be done by allowing measurements and performance
data of different vendors to share the same language so that the
network analysis and troubleshooting can be done easily.
However, most network vendors would prefer to have their
own self-optimization functionality so that they can easily
monitor their network based on their decision. As this issue
could not be avoided due to the choice of each network
operators, the designed self-organized control decision must be
reliable to operate in every language available.
B. Conflict between Parameters and Goals
Some of self-optimization use cases as described in [5]
share the same parameters for optimization purposes, but with
different goals. For example, to achieve better signal quality, a
user performs an immediate handover process to switch to
another cell that provides better signal level. Though, frequent
handovers will actually lead to fluctuations and signaling.
Meanwhile, in load balancing, number of handovers occurring
in a network should be at a minimum level to counter the
unbalanced traffic load in a cell. Therefore, a suitable
algorithm is required to combine both conflicting goals into a
target function based on the operator policy. For example, [9]
has proposed a new algorithm that achieved balanced traffic
load with minimum number of handovers.
In another example, SON’s ability to automatically power
on and power down the base station would actually be
beneficial for energy saving purpose. However, the abrupt
power level changes may affect the transmission performance
of several users especially for those who are located at the cell
edges. This is because inter-cell interference from the
neighboring base station cells might take place. In such
situation, operators should use other power saving strategy
such as dynamic or adaptive power control that considers
interference effects to the nearby possible affected users.
For that reason, what we can do is to coordinate together
the conflicting self-optimization goals due to the dependencies
on each other. This would overcome the conflicting issues. To
do so, a general priority shall be provided for each individual
use case. When the priority is set, the conflicting issues can be
solved simultaneously based on the priority to improve the
system.
C. Data Measurement and Processing
The most challenging issue in data measurement is to
decide what type of measurement data needs to be collected, at
what level the data needs to be collected, and appropriate
techniques to be used in collecting the data measurements. For
these reasons, a thorough understanding on the type of
networks/systems we are working with is essential. In
example, LTE-Advanced system is an OFDM-based
technology. Thereafter, specific parameters that are needed as
the measurement data can be easily determined. For instance,
signal-to-interference-and-noise-ratio (SINR) is an important
parameter in handling interference issues in both co-tier and
cross-tier networks. This parameter is assumed to be obtained
from the measuring reports that are feedback by the users to
their serving base stations. Soon after the required data were
successfully collected, the other challenge to be considered is
to design the best processing method that is efficient in
handling errors in the measurements’ reports.
D. Developing Algorithms
Trial and error method in altering real network parameters
and deploying new algorithms are prohibited as it is too risky,
requires extra effort, and may result undesirable performance.
Therefore, probabilistic approach is used in developing SON
algorithms since there are incomplete information and no
promising strategy to improve specific problem in a situation.
Though, finding an exact approach in handling a problem is a
challenging task as there are large parameters of space and
factors (e.g incomplete, delayed, and faulty feedbacks) that
need to be considered when designing an optimal algorithm.
To develop a more realistic algorithms, existing concepts
such as game theory, genetic algorithms, Q-learning, fuzzy
logic, heuristic and other tools are applied. These concepts
could improve the variability and the effectiveness of the
developed algorithms.
E. Evaluation Features
In radio performance evaluation, we can rely on simple
network topologies such as hexagonal grids and both
homogeneous users and traffic distributions. Although it
sounds easy, we cannot simply reuse the same performance of
evaluation assumptions of previous related designs. This is
because SON system design must be forwarded looking to
cater to future changes. Else, the design would not be reliable.
Also, very specific data bases of a particular area are
highly required towards planning real network tasks.
Moreover, trying to design self-optimization system on a real
network scenario is prohibited since it is a complex
investigation. In order to do so, we require expensive data
bases and complex simulators but yet the investigation will
still result in lack of generality. To counter the problem, what
we can do is by having several simulations at different
locations but of course, this method is time consuming and
requires high computational requirements.
For the time being, the best approach is to derive a feasible
method for evaluation to enhance the existed system
evaluation assumptions. For instance in simulating load
balancing, only simple models could be used to simulate
heterogeneous spatial and temporary UEs distributions.
III. SON IMPLEMENTATION IN PRACTICAL DEPLOYMENT
ASPECT
As discussed literally in [5], there are seven use cases in
SON self-optimization. However, authors are more interested
to discuss on inter-cell interference coordination in Het-Net.
This issue is the main challenge in ensuring a successful
deployment of such networks. In Het-Net, various types of
base stations are supported in the same spectrum, that is; a
macrocell is being overlaid by small cells such as picocells,
femtocells, and also the relay nodes. Since macrocell networks
and femtocell networks have same objective – to serve outdoor
users and indoor users respectively at the best point, they
cannot avoid from interfering each other when they are
operating under co-channel deployment. For that reason, many
researchers were focused on interference mitigation technique
in cross-tier cellular networks between macrocell and
femtocells such as in the existing literature [10–26].
In the early stage of intereference avoidance, most of the
research made use of frequency planning approach. This
approach includes the frequency orthogonality [10] where the
base stations of different cells are assigned with different set of
frequencies. Then, authors of [11] proposed a fractional
frequency reuse (FFR) scheme. Although this scheme allows
users at the center zone and cell-edge zone of a cell to use
different set of frequency, it occupies the same drawbacks as in
the previous technique. The drawbacks are the available
resource blocks for scheduling become limited and spectrum
are not used efficiently. Further, [12] introduced a dynamic
partitioning frequency reuse to increase the spectrum utilization
as compared to the frequency partitioning and FFR. Still, the
spectrum utilization is not fully utilized.
Besides taking advantage of frequency planning in avoiding
interference, other research works chose power control
technique to address the same problem. It can be achieved by
limiting the transmission power of a base station at a specific
power level. For example, power control proposed in [13]
solved the downlink interference problem in multi-tier
networks. The authors developed a macro-QoS (quality of
service) power control which considers interfered macro-users
in adjusting femtocell’s power level. Meanwhile, authors in
[14] introduced a downlink power control that is available in
both centralized and distributed solutions to solve the issue of
co-channel interference. Since centralized power control is
more subject to network delays and congestions, distributed
power control is more preferable for implementation and it
only requires local information of the network to control the
interference. Though power control might not be an optimal
way to achieve spectrum efficiency, fairness among cell-center
and cell-edge users is guaranteed. Other than that, there are
recent works that combined both frequency planning and
power control. For example, [15] came with a conclusion that
combining a fractional power control (FPC) with a dynamic
resource allocation offers attractive interference mitigation
while guarantees low complexity.
Since the number of femtocells increase and they are
installed anywhere and anytime by the users, femtocell
networks have been proposed to be self-organized entities as in
[12] and [16–18]. By occupying self-organizing features, each
femtocell would be able to reconfigure its radio parameter
based on the network environments in order to minimize
interference on the macrocell. The radio parameter adjustment
is possible since a self-organized femtocell has the ability to
scan and sense for available network resources (e.g. spectrum
opportunity and transmission power) through the air interface.
As pointed in [19], frequency assignment done in a self-
organized way gives better system performance then frequency
assignment done in a random manner. The organized way of
frequency assignment is achieved through information
exchange between the base stations or through the
measurement reports sent by the users. As presented in [20],
the combination of spectrum assignment and power control
transmission is achievable through self-organization concepts.
This work has proved that the role of SON has improved the
spectrum efficiency and reduced power consumption in a
network. In advance, authors in [21] use SON to perform an
optimal performance in enterprise femtocells. They concluded
that enterprise femtocell networks are more challenging from
femtocells networks that are being deployed in homes.
Self-organizing femtocell networks can be further improved
by utilizing machine learning technique such as Reinforcement
Learning (RL). This learning process allows the self-learning
entity (femto-base stations in this case) to tune its parameter
without the need to know about other entities’ actions. Existing
research on RL (focusing specifically on Q-learning based
algorithm) in femtocell networks have been carried out in [16]
and [22–25]. For instance, no communication among
femtocells is considered in [22]. Thus, macro-base station
would send a message telling the interfering femtocells to
adjust their power control and transmission channel in order to
alleviate interference in its network. On the other hand, [23]
focused on interference mitigation technique via power
allocation. The power allocation is performed in distributed
manner in each femtocell by assigning different power levels to
the assigned resources with regards to the measured
interference levels at that time. Meanwhile, as proposed in
[24], each cell decides its frequency resource for transmission
in a way that SINR of each cell improves.
IV. FUTURE TRENDS
The existence of huge number of femtocells in LTE-
Advanced networks introduces severe interference among
themselves and to macro-user network. Therefore, finding the
best approach to alleviate co-channel interference is still an
open research and there are many opportunities for future
innovative works toward overcoming these issues. The key
solution in mitigating interference is through frequency
planning/allocation along with the power control. It would be
in order, easier, less complicated and invulnerable to errors
when self-organizing functionalities are applied to interference
mitigation/avoidance approach. The functionalities help the
entity to integrate itself into the network through self-
configuration mechanism, become sensitive and learn about
the surrounding environment and react with the changes
accordingly by tuning the network parameters.
Additionally, recent researchers prefer to apply RL
together with self-organized interference management
techniques. The application of this machine learning allows
reduction in overhead signaling. This is because, a femtocell
learns and adapts with the changes in the environment without
depending on the information exchange between the
neighboring femtocells. Other than that, distributed network
planning is feasible through this RL-based approach.
An alternative way is to use advanced interference
management such as Coordinated Multi Points (CoMP).
CoMP is one of the key technologies in LTE-Advanced
systems and promises an efficient interference management
technique. In the meantime, this technology would be able to
provide robust performance, improve inter-cell fairness issue,
and enable gains in spectral efficiency. For instance, a novel
inter-cell interference mitigation scheme using CoMP was
proposed by authors in [26]. It was proved that the proposed
scheme has increased the cell-edge user performance and
maintained the overall system performance.
V. CONCLUSIONS
In this paper, we provide an overview on SON in advanced
wireless networks such as Long Term Evolution (LTE) and
LTE-Advanced systems. Although SON promises huge
benefits towards having reliable and optimum mobile
networks, the benefits come with challenges in making them
reality. Thus, aspects such as reliability; conflict between
parameters and goals; data measurement and processing;
algorithms development; and evaluation features must be
taken into consideration before implementing self-
organization concepts in practical. Other than that, we present
and discuss the examples of SON implementation in femtocell
networks to mitigate inter-cell interference to the macrocell.
The practical deployments show that SON approach is very
effective for dense network deployment such as Het-Net in
mitigating interference between different cells and consuming
energy efficiently towards achieving green networks.
ACKNOWLEDGMENT
The authors would like to thank Ministry of Higher
Education of Malaysia (MOHE), Research Management Centre
(RMC) of Universiti Teknologi Malaysia, and UTM-MIMOS
Centre of Excellence and for their full support and advice in
realizing this research project. Also, thanks to all anonymous
reviewers for their invaluable comments and the guest editors
for handling the review for this paper. The work is financed
under grant of Q.J130000.7123.01H99.
REFERENCES
[1] C. Prehofer and C. Bettstetter, "Self-organization in communication
networks: Principles and design paradigms," IEEE Communications
Magazine, vol. 43, pp. 78-85, 2005.
[2] R. Jagadish, “Self Organized Radio Access Wireless Networks,” L&T
Infotech Proprietary.
[3] Texas Instruments, A. Gatherer, P. Dent, S. Bhadra, and R. Vedantham,
“Self-optimizing networks (SON): doing more with less,” White Paper,
2009.
[4] 3GPP Work Items on Self-Organizing Networks v.0.0.6, Oct. 2010.
[5] 3GPP TR 36.902 v9.2.0, “Evolved Universal Terrestrial Radio Access
Network (E-UTRAN); Self-Configuring and Self-Optimizing Network
(SON) Use Cases and Solutions (Release 9),” 2010.
[6] N. Marchetti, et al., "Self-organizing networks: State-of-the-art,
challenges and perspectives," 8th International Conference on
Communications (COMM), 2010.
[7] M. Döttling and I. Viering, "Challenges in mobile network operation:
Towards self-optimizing networks," IEEE International Conference on
Acoustics, Speech, and Signal Processing (ICASSP), 2009.
[8] NEC Corporation, “Self organizing network: NEC’s proposals for next
generation radio network management,” White Paper, 2009.
[9] H. Hu, et al., "Self-configuration and self-optimization for LTE
networks," IEEE Communications Magazine, vol. 48, pp. 94-100, 2010.
[10] C. Kyongkuk, et al., "Resource alloation for orthogonal and co-channel
femtocells in a hierarchical cell structure," IEEE 13th International
Symposium on Consumer Electronics (ISCE), 2009.
[11] P. Lee, et al., "Interference management in LTE femtocell systems using
fractional frequency reuse," 12th International Conference on Advanced
Communication Technology: ICT for Green Growth and Sustainable
Development (ICACT), 2010.
[12] M. Z. Chowdhury, et al., "Interference mitigation using dynamic
frequency re-use for dense femtocell network architectures," 2nd
International Conference on Ubiquitous and Future Networks (ICUFN),
2010.
[13] S. P. Yeh, et al., "Power control based interference mitigation in multi-
tier networks," IEEE Globecom Workshops (GC), 2010.
[14] X. Li, et al., "Downlink power control in co-channel macrocell femtocell
overlay," 43rd Annual Conference on Information Sciences and Systems
(CISS), 2009.
[15] G. W. O. Da Costa, et al., "Interference mitigation in cognitive
femtocells," IEEE Globecom Workshops (GC), 2010.
[16] A. Galindo Serrano, et al., "BeFEMTO's self-organized and docitive
femtocells," 2010 Future Network and Mobile Summit, 2010.
[17] A. De Domenico and E. C. Strinati, "A radio resource management
scheduling algorithm for self-organizing femtocells," IEEE 21st
International Symposium on Personal, Indoor and Mobile Radio
Communications Workshops (PIMRC), 2010.
[18] C. H. M. De Lima, et al., "Interference management for self-organized
femtocells towards green networks," IEEE 21st International
Symposium on Personal, Indoor and Mobile Radio Communications
Workshops (PIMRC), 2010.
[19] D. López-Pérez, et al., "OFDMA femtocells: A self-organizing approach
for frequency assignment," IEEE 20th Personal, Indoor and Mobile
Radio Communications Symposium (PIMRC), 2009.
[20] F. Bernardo, et al., "Self-optimization of spectrum assignment and
transmission power in OFDMA femtocells," 6th Advanced International
Conference on Telecommunications (AICT), 2010.
[21] G. De La Roche, et al., "Self-organization for LTE enterprise
femtocells," IEEE Globecom Workshops (GC), 2010.
[22] M. Bennis and D. Niyato, "A Q-learning based approach to interference
avoidance in self-organized femtocell networks," IEEE Globecom
Workshops (GC), 2010.
[23] A. Galindo-Serrano and L. Giupponi, "Distributed Q-learning for
interference control in OFDMA-based femtocell networks," IEEE 71st
Vehicular Technology Conference (VTC), 2010.
[24] F. Bernardo, et al., "Intercell Interference Management in OFDMA
Networks: A Decentralized Approach Based on Reinforcement
Learning," IEEE Transactions on Systems, Man and Cybernetics Part C:
Applications and Reviews, 2011.
[25] Z. Feng, et al., "Reinforcement learning based dynamic network self-
optimization for heterogeneous networks," IEEE Pacific Rim
Conference on Communications, Computers and Signal Processing
(PACRIM), 2009.
[26] Z. Wang, et al., "A novel Inter-Cell Interference mitigation scheme for
downlink of LTE-Advanced systems," 3rd IEEE International
Conference on Broadband Network and Multimedia Technology (IC-
BNMT), 2010.