Abstract:
Information regarding licensed primary user (PU) space positioning can allow enabling
of several important attributes in cognitive radio (CR) networks such as intelligent
location-aware routing, improved spatio-temporal sensing, along with aiding spectrum
policy enforcement. In this work, the issue of PU location estimation in presence of
CRs which are outlier is dealt with. This is an noteworthy problem to address practically
as in many real-world scenarios the CRs reports unreliable information. Therefore,
rstly the accuracy that PU localization algorithms can achieve by jointly utilizing direction
of arrival (DoA) and received signal strength (RSS) measurements is considered
by evaluation of Cramer-Rao Bound (CRB). In past research, CRB for DoA-only and
RSS-only localizationialgorithms are evaluated separately and estimationierrorivariance
of DoA is assumed to be independent of RSS. In this work, for jointiRSS and DoA-based
PU localizationialgorithms, CRB is evaluated which is based on mathematical model in
which DoA is dependent on RSS. The bound is then used in futher work to examine the
performance of PU localization algorithms and impact of number of CRs is discussed.
CRB for uniformirandomiCR deployment is also derived and studies are performed to
nd out number of CRs tightly approximate integration of CRB for xed CR placement
by asymptotic CRB.
Following that statistics techniques are applied on squared range measurements and
two di erent methods are implemented for solving the task of PU localization in presence
of outlying CRs. The rst approach is e cient in terms of computational complexity, but
only objective convergence is guaranteed theoretically in that approach. Contrary to that,
whole-sequence convergence is established for second method . In order to take bene ts of
both the approaches, a hybrid algorithm is developed by integrating both the approaches
that o ers computational e ciency along with whole-sequence convergence.Simulations
show that robust methods meet the CRB for large number of CRs. For small number of
CR measurements, the implemented robust methods does not achieve CRB but performs
better than other localization methods implemented in this work.
