Kalman Modeling of Multipath Wireless Channel

This paper evaluates the performance of linear Kalman innovation models in the observable block companion form, for describing the dynamics of measured received signal strength indicator (RSSI) in multipath wireless channels. First, measurements are conducted inside a single-story housing of complex geometry, to generate a data set consisting of discrete samples of the RSSI of the signal between a mobile phone configured as a 2.4 GHz router, and a spectrum analyzer. Then, Kalman models are constructed from linear autoregressive moving average (ARMA) models, identified using extended least squares with different number of previous values of the RSSI and the estimation error. Results show that the linear Kalman models have a satisfactory performance in filtering out the noise, thus generating innovation errors with a degree of randomness sufficiently high to be exploited in the generation of secret keys for secure communication in multipath channels.

Major Advisor: Huaping Liu
Committee: Bella Bose
Committee: Eduardo Cotilla-Sanchez
Committee: Tejasvi Anand
GCR: David Porter