Abstract:
RF localization has gained prominence because of its potential for supporting various
position based applications. Passive RF Localization based on Received Signal Strength
Indicator (RSSI) uses the strength of received signal from a target by passive listening to
infer the range, which is subsequently used for position estimation. The thesis undertakes
a study of localization techniques and addresses the problem of accuracy of position
estimation. State space model developed for localization is nonlinear and hence does not
have a closed form solution. Posterior density for state vector has been derived and
simulated using a variant of Kalman Filter and Monte Carlo methods to obtain respective
sub-optimal solutions. Least Squared Error approach tries to obtain an estimate that
minimizes the squared error whereas and does not reveal any statistical information about
the target location. Extended Kalman filter approach tries to estimate the posterior
density of target employing approximations of Gaussian state probability distribution and
linear state space model and observed to provide better results compared to that of Least
Squared Error approach. As the localization model is nonlinear, Extended Kalman filter
approximates it with a linear one by employing Taylor series approximation and if the
nonlinearity is severe the accuracy of the algorithm suffers. Particle filter approach also
tries to estimate the state posterior density with no restrictions and hence is applicable for
any generalized system. In this approach probability density function is approximated
using a weighted set of particles drawn using Monte Carlo methods and will enable in
computing the all the moments of distribution. Recursive Least Squared Error, Extended
Kalman and Sampling Importance Resampling Particle Filtering algorithms are designed
for localization and their performances are compared. The performance of Particle filter
using Sampling Importance Resampling algorithm is found to be superior to that of
Recursive Least Squared Error approach and Extended Kalman filter.
