Department of Electrical and Computer Engineering

Robust filtering and control strategies with applications to wireless networks, Dr. Ananth Subramanian, PhD. UCLA

Robust filtering and control strategies with applications to wireless networks 

Dr. Ananth Subramanian
PhD. UCLA

Time: 11-12, Feb 28, 2005
Place: Coover 2222

Abstract: Given a Gauss-Markov signal model that consists of a linear dynamical system driven by a white noise process, the Kalman filter provides the optimal linear least-mean-squares estimator for the system states. An important premise in Kalman filter theory is that the underlying state-space model is accurate. When this assumption is violated, the performance of the filter can deteriorate appreciably.  This filter sensitivity to modeling errors demands and motivates the development of robust state-space filters; robust in the sense that they attempt to limit, in certain ways, the effect of model uncertainties on the overall filter performance.

In this talk, we develop robust filtering schemes that address some challenges posed by un-modeled dynamics, such as uncertainties arising from mixed deterministic and stochastic sources as well as state-delayed dynamics. We also develop certain multi-objective filters that are exponentially stable, guarantee bounded error covariance matrix, and meet an H-infinity performance measure. The tools employed for robust filter design are then extended to develop robust control strategies applied to problems in wireless networks. Specifically, we will show how to develop a robust filtering and control algorithm for joint rate and power control in wireless networks. Power consumption normally is a key limiting factor in the performance of wireless networks. This limitation is further compounded by the fact that nodes in a network need to cater to desired data rates and link congestion levels (say, as in TCP over wireless), which in turn require the SNR level, and consequently the power level, to be above certain values.  From a system-theoretic perspective, the problem requires that we deal with state-delayed models with and without uncertain dynamics. As a result of the analysis, we will end up with a joint rate and power control algorithm that minimizes a bound on the error variance between the desired and actual signal-to-interference ratios (SIR).
In addition, we will address robust schemes for applications like sensor node localization. We will propose a combination of robust filter and least squares strategy for localization of nodes. As such, the talk focuses both on the theoretical underpinnings of robust filtering and control algorithms and on their applications in the context of wireless networks