IEEE Computational Intelligence Magazine - May 2018 - 59

margin and the TTT taken by each
SON function.
B. SOn Conflict

In order to explain the SON conflict, we
implement MLB and MRO SON functions in a distributed manner. As previously introduced, both of them aim at
adjusting the CIO, hysteresis and the
TTT handover parameters for different
purposes.The goal of MLB is to optimize
the network quality of service (QoS) by
evenly distributing the load among the
different cells. On the other hand, the
MRO aims at decreasing the number of
radio link failures (RLFs) caused by tooearly or too-late handovers.
Figure 6 shows that the two SON
functions are implemented in a distributed manner and concurrently executed in
a per-cell manner. In order to analyze the
MLB-MRO SON conflict, we assume
that the serving cell (cell A) is overloaded
and its neighbor (cell B) has a lower load.
Therefore, cell A chooses its neighbor
cell B to balance the load, and the action
request is to adjust the CIO, the hysteresis
and TTT handover parameters. This way,
the condition in Event A3 can be met by
UEs close to the cell edge. Hence, several
handovers will happen from cell A to
cell B, thus transferring part of the load to

The proposed scheme, BDA-NSGA-II, allows to
obtain a set of network configurations that improves
its performance when conflicting optimization
functionalities are present.
cell B . This results in too-early handovers
of UEs in cell A causing an increase in the
RLFs rates and ping-pong effects. As a
consequence, the MRO SON function
detects these metrics and tries to reduce
them by requesting new changes in CIO,
the hysteresis and TTT parameters. Since
cell A is still overloaded, the MLB changes
the handover parameters again and the
MRO will also change them in the
opposite direction, thus yielding an endless loop.
C. Performance indicators

Besides the UE measurement reports,
which contain RSRP and RSRQ values coming from the serving and the
neighboring cells, the data about the
performance of each UE is also collected once it performs a handover. These
metrics are chosen based on the objective of each SON function.
On the one hand, the main objective
of MLB is to improve end-user experi-

ence and achieve higher system capacity
by distributing user traffic across system
radio resources. As a consequence, the
load of a cell is measured in terms of the
average physical resource blocks (PRBs)
that can be allocated to the users and the
average signal to interference plus noise
ratio (SINR) of each cell.The number of
bits at the physical layer, referred to as
the transport block size (TBS), is chosen
taking into account the data that need to
be transmitted by a UE. The media
access control (MAC) has to first decide
on the modulation scheme that can be
scheduled to the user and then check the
physical resource grid for the availability
of the resource blocks. Given this, the
MAC can decide upon the modulation
and coding scheme index and its TBS
index taken from [25], and then decide
the number of PRBs that can be allocated to the user (i.e., users are allocated a
specific number of subcarriers for a predetermined amount of time).

Hysteresis - CIO_MRO
Hysteresis - CIO
Hysteresis - CIO_MLB
Handover Start

Handover Start

Hysteresis
TTT
Condition Met
78 Neighbor Cell
Becomes Better
77 than Serving Cell
Hysteresis
76

Hysteresis
TTT
Condition Met
Neighbor Cell
Becomes Better
than Serving Cell

Neighbor Cell

75
CIO

Serving Cell

78
77

Neighbor Cell

76
75

Serving Cell

74
73

73
72
36

RSRP (dBm)

72
37

39
40
Time (s)
(a)

39
40
Time (s)
(b)

Figure 5 Effect of the handover parameters. The figure on the left side depicts the perceived RSRP of serving cell and a neighboring cell by a
UE, which crosses the border of the cells. By default, the algorithm uses a hysteresis margin of 3.0 dB and a TTT of 256 ms. On the right, the
MLB and MRO tune these values in order to achieve their own goals.

may 2018 | IEEE ComputatIonal IntEllIgEnCE magazInE

Table of Contents for the Digital Edition of IEEE Computational Intelligence Magazine - May 2018