Self-optimizing adaptive transmission mode selection
for LTE-WLAN aggregation
Irina Balan, Eva Perez, Bernhard Wegmann
Nokia Bell Labs, Munich, Germany
{irina.balan, eva.perez,
bernhard.wegmann}@nokia-bell-labs.com
Daniela Laselva
Nokia Bell Labs, Aalborg, Denmark
daniela.laselva@ nokia-bell-labs.com
Abstract— As licensed spectrum is scarce and users have ever
increasing throughput and data volume requirements, mobile
network operators (MNOs) are looking for innovative ways for
using the unlicensed spectrum. One such solution is LTE-WLAN
aggregation (LWA) where a LTE radio bearer can be served by
making use of both LTE and WLAN resources in a flexible
manner. This paper proposes a self-optimizing algorithm that
tunes the aggregation mode and controls how data is split between
LTE and WLAN. Results show that the proposed algorithm is able
to derive the optimal control parameter settings per cell under
different load conditions. As a result, improved average and 5th
percentile UE throughput values are observed along with a more
even throughput distribution among LWA and non-LWA users.
The performance of individual cells in the scenario is also leveled
out thanks to the SON algorithm.
Keywords—SON; LTE; WLAN; LWA; adaptive transmission
mode selection; bearers split; bearer switch
I. INTRODUCTION
The ever increasing traffic demand by users is pushing
mobile network operators not only to invest in new radio access
technologies but also to seek the use of unlicensed spectrum in
order to increase their network capacity and coverage.
To this end, several options have been standardized in 3GPP.
In Release 13, Licensed-Assisted Access (LAA), LTE is adapted
for use in the unlicensed spectrum. The Release 12/13 features
of LTE-WLAN Interworking (LWI), defined RAN
assisted/controlled traffic offloading to the WLAN, which
releases LTE resources. In Release 13, LTE-WLAN
Aggregation (LWA) defined Packet Data Convergence Protocol
(PDCP) level aggregation of the WLAN and LTE resources; via
LWA, users could make use of both radios simultaneously for
the same radio bearer, similar to LTE Dual Connectivity (DC)
[1]. Additionally, in Release 13 the LTE integration with legacy
WLAN Access Points (APs) is defined by switching a radio
bearer to an IPsec (Internet Protocol Security) tunnel, via
WLAN, between the eNB (Evolved NodeB) and UE (LWIP).
In LWA, the eNB, based on UE measurement reports and
WLAN network information (received via a newly defined Xw
interface between an eNB and as a termination of the 3GPP
interfaces (C-plane and U-plane) on the WLAN side, namely the
WLAN Termination or WT), determines if aggregation should
be used by the UE and how data is to be split between the two
radios.
Numerous studies have been carried out to test these various
approaches of using the unlicensed spectrum or to compare their
performance. In [2], the performance of offloading approach
(LWI) has been investigated with respect to the UE throughput
performance considering different user distributions among the
two layers, under different LTE load levels and measurement
errors in the UE. [3] and [4] are comparative studies of LWI,
LAA and LWA. LWA outperforms LWI as the users are
summing up the throughputs over two links as opposed to use
only one of them at the time, but this also means that one user is
contributing to load in both systems. LAA brings the advantages
of LTE (e.g. higher spectral efficiency) to the unlicensed
spectrum and can coexist fairly with WLAN.
Although all of these tight interworking solutions show
promising performance gains, they are highly sensitive to the
static parameter settings and their gain over LTE only or LWI
varies with the load of the system and the deployment, e.g. the
gain drops when load increases, similarly to LTE DC.
Also, due to the dynamics of the investigated systems and
scenarios, looking at average performance can be misleading.
Cell specific performance may vary based on a multitude of
factors (network deployment, UE mobility, cell load, etc) and
diverge from the average performance. An algorithm that would
deal with any combination of factors and tune the control
parameters independently per cell, in a self-optimizing fashion
would be easy to use by the operator and would bring
performance gains.
Self-Organizing Network (SON) algorithms are well known
to outperform their static setting counterparts in a dynamic
system. An example is given in [5] that used a SON algorithm
for optimizing autonomously the access network selection
between the LTE and WLAN systems (LWI). Results show a
significant improvement in the median and 5th percentile user
session throughput compared to the ‘connect to WLAN when
available’ reference case.
However, to the best of the author’s knowledge, there is no
prior attempt of designing a SON algorithm for combining the
aggregation and offload in the context of LWA. The latter can
be viewed as a special case of aggregation where only WLAN
link is used. First, an algorithm that determines the adaptive use
of WLAN resources and LTE resources depending on the radio
conditions of the users in both systems was designed under the
name of adaptive transmission mode selection for LWA. The
primary goal of this algorithm was leveling out the performance
2016 IEEE 27th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC): Workshop:
6th International Workshop on Self-Organizing Networks (IWSON)
978-1-5090-3254-9/16/$31.00 ©2016 IEEE