| Summary: | 碩士 === 國立臺灣科技大學 === 機械工程系 === 98 === The Kalman filter is a recursive optimal filter that estimates the
States of a linear dynamics system, efficiently from a series of noisy
measurement. An ideal Kalman filter can lead to an optimal filter when
dynamic model of system is completely known, linear and noise must
be white with zero mean. However in reality, linear systems do no really
exist, that means system’s state space equation and measurement
equation are all nonlinear. In this case, Kalman filter can not suffice us
for precise positioning. If we want to resolve this kind of problem,
nonlinear filter estimation must to be used for higher localization
accuracy.
In this thesis, three methods of nonlinear estimation algorithm
including Extended/Unscented Kalman Filter and Particle Filter
are applied for simulating WSN/DR integration positioning in dynamic
indoor environment . In the framework of the system, two ZigBees and
two inertial navigation sensors were used to obtain measurement
informations for estimating the position of moving body, parameters
of ZigBee’s RSSI and other relating states.
Simulation results show that Particle Filter can estimate the states
of the system more accurate than EKF and UKF, efficiently reducing
the effect of unstable RSSI measurements in the dynamic indoor
environment.
|
|---|
