Networking Research 

Prachee Sharma  

BE, Ravishankar University, Raipur, India, 1991 

MS, Computer Science, University of Massachusetts Lowell, 1998

Sc.D, Computer Science, University of Massachusetts Lowell, 2003

Work: Research Associate,  Indian Institute of Science,  Bangalore, India 

Thesis Title: Predictive Models for Wireless Fading Channels 

Thesis (pdf)


This thesis presents methods for improving the performance of wireless networks through the modeling and prediction of time-varying multipath channels. The Rayleigh fading channel is characterized using first and second-order autoregressive (AR) time-series models. The AR processes model the channel variations at the time-scale of the characteristic Doppler frequency. Small time-scale variations are captured using linear interpolation of the AR model predictions. For defined error performance metrics, the fading signal is characterized using a state-space model that partitions the continuous amplitude variations into error and error-free states. The model parameters and state thresholds are derived as a function of Doppler frequency, signal-to-noise ratio and specified probability of error. The aforementioned model is applied for estimating the probability of error as a function of transmission block size. The second-order AR model captures the pseudo-periodic behavior of the Rayleigh channel and produces accurate estimates of block error probabilities relative to the first-order AR process. The fading channel model is applied for both flat and frequency-selective channels in the design of a channel estimator and predictor. A Kalman-filter is designed to forecast the expected channel conditions and the predictions are applied for error control. The performance of the proposed error control approach is compared to a decision feedback equalizer (DFE). The model-based prediction shows at least a 25% improvement over the DFE.