An expert system for real-time traffic management in wireless local area networks


An expert system for real-time traffic management in wireless local area
networks

T. Frantti a,⇑, M. Majanen b
a Renesas Mobile Europe Ltd., Elektroniikkatie 10, FI-90570 Oulu, Finland
b VTT Technical Research Centre of Finland, Finland

a r t i c l e i n f o

Keywords:
Expert system
Fuzzy control
Flow control
Offloading
Real-time traffic
WLAN

a b s t r a c t

The paper explores delay-based congestion and flow control and the offloading of real-time traffic from
wireless local area networks (WLANs) to mobile cellular networks (MCNs) in multihomed devices. The
control system developed is based on an embedded hierarchical expert system. It adjusts transceivers’
traffic flow(s) for prevailing network conditions to achieve application-dependent delay and throughput
limits. In wireless networks, delay and throughput depend on the packet size, packet transmission inter-
val, and node connection density. Therefore, the controller on the destination node monitors average
one-way delay and the change of one-way delay of the incoming traffic. On this basis, it adjusts the
packet size and transmission interval of the source node by transmitting a control command to the
source. If the prevailing level of traffic in the network exceeds its capacity despite of the control actions
taken, devices prepare for developed asynchronous offloading of traffic to another access network.

The control model was validated via simulation of Voice over Internet Protocol (VoIP) traffic in the
OMNeT++ network simulator. The results demonstrate that the expert system developed is able to reg-
ulate packet sizes to match the prevailing application-dependent optimum and transfer traffic to another
network if the network exceed its capacity no matter the control actions taken. Although this work is
motivated mainly by issues of congestion and flow control of WLAN systems and the simulations and
results were prepared for the IEEE 802.11b system, the approach and techniques are not limited to these
systems, but they are applicable for other packet switched access networks (PSANs), too.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

With the everyday use of cellular networks, extensive mobility
during communication became self-evident reality for end users. A
mobility requirement soon appeared also for Internet-based com-
munication, and currently mobile phone networks are moving
with Long Term Evolution (LTE) toward an Internet Protocol (IP)
infrastructure with the ultimate aim of offering almost all the ser-
vices provided by current second- and third-generation (2G/3G)
cellular networks through an all-IP network (see ITU-T (2004a)
and ITU-T (2004b) for more details), although Transmission Con-
trol Protocol/Internet Protocol (TCP/IP) networks were not origi-
nally designed for mobile use. For example, devices on the
Internet are identified by an IP address, which has a dual role: it
serves as an identifier and as a locator of the networked device.
If a device changes its point of attachment, such as wireless access

network, its IP address too may change, which means that other
devices in the network need means to access the new address if
they are to reach the device at its new location. The change in IP
address causes the upper-layer protocol (at the layer above the
network layer) connection to break, which is problematic for appli-
cations with more persistent connections or applications requiring
registration of an IP address.

Today, mobility in IP networks relies on wireless access technol-
ogies, and the trend is toward equipping mobile devices, such as
smartphones, with multiple network interfaces. The various inter-
faces are, in the general case, operated by different Internet service
providers (ISPs), for devices’ improved resilience via an opportu-
nity to connect to the Internet through at least one of the access
technologies. For end users, these technologies vary mainly in their
coverage area and performance.

A networked mobile device can use the available interfaces
either one at a time or several simultaneously. The device using
its multiple network interfaces simultaneously is a special case of
node multihoming. A multihomed device has multiple IP ad-
dresses, assigned to the same network interface or different ones.

http://dx.doi.org/10.1016/j.eswa.2014.01.024
0957-4174/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +358 40 547 0819.
E-mail addresses: tfrantti@ee.oulu.fi (T. Frantti), mikko.majanen@vtt.fi

(M. Majanen).

Expert Systems with Applications 41 (2014) 4996–5008

Contents lists available at ScienceDirect

Expert Systems with Applications

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / e s w a

http://crossmark.crossref.org/dialog/?doi=10.1016/j.eswa.2014.01.024&domain=pdf
http://dx.doi.org/10.1016/j.eswa.2014.01.024
mailto:tfrantti@ee.oulu.fi
mailto:mikko.majanen@vtt.fi
http://dx.doi.org/10.1016/j.eswa.2014.01.024
http://www.sciencedirect.com/science/journal/09574174
http://www.elsevier.com/locate/eswa


Node multihoming differs from site multihoming, which can be de-
fined as an edge network configuration that has more than one ser-
vice provider but does not provide transit communication between
them; see the work of Baker (2011). In site multihoming, end users
do not need to manage their devices in any way. When a net-
worked device is multihomed, there are multiple paths between
the source and destination devices. The paths may differ in price,
data rate, latency, jitter, and packet loss. Node multihoming, with
parallel connections, enables new kinds of traffic engineering tech-
niques, whilst site multihoming is intended mainly to increase the
reliability of the Internet connection.

In this paper, we explore delay-based congestion and flow con-
trol and offloading of real-time traffic from wireless local area net-
works (WLANs) to cellular networks in multihomed mobile
devices. In WLANs, delay and throughput depend on the packet
size, packet transmission interval, and connection density. There-
fore, we developed and applied control systems to adjust trans-
ceivers’ packet sizes for prevailing network conditions to achieve
application-dependent delay and throughput limits for real-time
traffic and to avoid unnecessary offloading. If the prevailing level
of traffic in networks exceeds capacity regardless of the control ac-
tions, devices prepare to perform asynchronous offloading of traffic
to another access network. In our work, the research assumptions
can now be presented as: ‘‘Multihomed and multiradio (WLAN and
cellular) devices operating in WLAN and running real-time applica-
tions try to maximize network capacity by packet size optimization.
When the performance threshold is reached, connections are asyn-
chronously offloaded to cellular networks.’’ By real-time traffic we
mean Voice over Internet Protocol (VoIP), video calls, and interac-
tive games. With the assumptions above, the research question for
the work can be formulated as follows: ‘‘When should we begin to
transfer traffic of real-time applications to another (cellular) network
as the WLAN’s traffic and number of users increase, and how can we
do that to avoid unnecessary offloading?’’

The control model was validated through simulation of VoIP
traffic in the OMNeT++ network simulator. Even if this work is ad-
dressed mainly to congestion and flow control of WLAN systems,
the approach and the techniques are not limited to these systems.
They are applicable also for other packet switched access networks
(PSANs).

The rest of the paper is organized as follows. Section 2 presents
a review of the literature on packet size control in WLANs, multih-
omed path management, and network selection. Section 3 focuses
on multihome- and offloading-related standardization activities.
Section 4 summarizes the hierarchical decision making system
developed, which features a packet size control unit and real-time
offloadig of traffic from WLANs to mobile cellular networks unit.
Section 5 describes the simulation model, and Section 6 presents
the simulation results. Section 7 delineates our future research. Fi-
nally conclusions are drawn in Section 8.

2. Literature review

2.1. Packet size optimization in WLANs

Korhonen and Wang (2005) have studied the effect of packet
size on loss rate and delay in an IEEE 802.11-based WLAN. The
analysis shows that there is a straightforward connection between
packet size, bit error characteristics, and observed delay character-
istics. In general, it is evident throughout the literature that the
performance of wireless networking is sensitive to packet size
and that significant performance improvements are obtained if a
‘‘good’’ packet size is used. For example, Bakshi, Krishna, Vaidya,
and Pradhan (1997) show this for TCP traffic over a wireless net-
work. Chee and David (1989), Lettieri and Srivastava (1998), and

Chien et al. (1999) all have studied the relationship between frame
length and throughput, but they do not propose any precise meth-
od for dynamic control of frame length to maximize throughput.
Smadi and Szabados (2006) focus on optimization of packet size
in error recovery but do not consider the optimal packet size for
performance optimization. Sheu, Lee, Chen, Yu, and Huang (2000)
present a fuzzy packet length controller (PLFC) for improving the
performance of WLANs suffering from interference from a micro-
wave oven. It was demonstrated that the PLFC improves the
throughput of User Datagram Protocol (UDP) traffic from that with
fixed-length packets, but the authors did not consider performance
improvements when the number of users and the amount of traffic
are increased. Sankarasubramaniam, Akyildiz, and McLaughlin
(2003) have studied packet size optimization for energy efficiency,
and Younis, Farrag, and D’Amico (2009) consider it for security and
throughput, but their solutions are statistical in nature, meaning
that the packet size is optimized beforehand. In the most recent
of our publications (Frantti & Majanen, 2011), we presented and
compared proportional-integral-derivative (PID) and fuzzy control
systems to achieve maximum throughput and minimal delay by
adjusting the packet sizes of UDP-based uni- or bi-directional
real-time traffic in WLANs in line with prevailing channel condi-
tions. The aim with the hierarchical expert system developed for
this paper is to reach application-dependent delay and throughput
limits for the maximum number of such real-time connections as
VoIP calls and, if necessary, offload real-time connections to an-
other network interface in a controlled manner.

2.2. Multihomed path management

In node multihoming, the end nodes are responsible for paths’
management. Fekete (2010) notes that multihoming can be used
to enable more advanced traffic engineering or load sharing tech-
niques, such as load balancing and load spreading, by distributing
traffic over multiple interfaces and addresses. In load balancing,
different traffic flows are sent on different paths. In load spreading,
packets belonging to the same flow are sent through different
paths. Kandula, Lin, Badirkhanli, and Katab (2008), Singh, Alpcan,
Agrawal, and Sharma (2010), and Yao, Kanhere, and Hassan
(2009) use quality of service (QoS) parameters for balancing the
traffic load over the available network interfaces. In host-centric
traffic engineering, this is done by selecting the proper source IP
address for outgoing packets and notifying peer nodes about the
IP address for incoming packets. In network-centric traffic engi-
neering, this is done within the network through tuning of the
routing protocols (see Schiller (2005)).

2.3. Multihomed interface selection

Common approaches to network selection are based on esti-
mated QoS parameters for each available network. Adamopoulou,
Demestichas, Koutsorodi, and Theologou (2005), Bari and Leung
(2007), Psaras and Mamatas (2011), Song and Jamalipour (2005),
Wang, Katz, and Giese (1999), Wilson, Lenaghan, and Malyan
(2005), Xing and Venkatasubramanian (2005) and Yahiya and Cha-
ouchi (2009) consider network QoS parameters such as delay and
capacity. In the literature, researchers have also proposed more
user-centric criteria, among them the power consumption of de-
vices (see Yahiya & Chaouchi (2009) and Petander (2009)) or the
cost of network use (see Adamopoulou et al. (2005), Bari & Leung
(2007), Wilson et al. (2005) and Alkhawlani & Ayesh (2008)). In
some of the aforementioned references, including those of Adamo-
poulou et al. (2005), Song and Jamalipour (2005), Wang et al.
(1999), and Alkhawlani and Ayesh (2008), a user is expected to
supply policies and preferences. The approach of Alkhawlani and
Ayesh (2008) proposes a combination of fuzzy logic and genetic

T. Frantti, M. Majanen / Expert Systems with Applications 41 (2014) 4996–5008 4997



https://isiarticles.com/article/52584