Abstract:
Often a cellular user is connected with a single cellular service provider via a subscriber
identity module (SIM). Such a cellular user is totally dependent on the particular access network
for cellular communication. This means that the user has to face different problems, such as signal
degradation, low bandwidth, connection failure, poor voice quality, unsecure network and high
power consumptions, which are associated with the access network. All these problems can be
reduced if the user has the choice to connect with multiple cellular networks. This gives rise to the
problem of network selection in a heterogeneous-cellular network environment (HCNE). The goal
is to identify network selection criteria that can classify all the access networks according to their
performance and select the highest-ranking network for the user. In this thesis, we discuss the
utility function based network selection schemes for HCNE. The utility function includes different
criteria related to the network performance that are each weighted by the user preferences.
Different single-criterion utility functions and multi-criterion utility functions have been proposed
in the literature. Recently multiplicative utility function has been proposed for network selection
that overcomes the limitation of inter-independency of performance criteria in the well-used
additive utility function. However we observed that the multiplicative utility function increases the
network elimination factor. Therefore a new exponential multi-criteria utility function has been
proposed in this thesis for network selection and it is showed that the proposed utility function is
monotonically related to the average of additive and multiplicative utility functions. This means
that the proposed function not only increases inter-dependency but also reduces the network
elimination factor. To highlight the importance of our network selection criteria, we consider a
x
specific HCNE with different performance criteria including upward and downward criterion and
compare their classification with additive and multiplicative utility criteria. The results are
obtained in MATLAB and show the effectiveness of our proposed method.
