Robust time-varying fading channel estimation with risk sensitive filter under parameter and process noise uncertainty

Abstract

An alternative to Kalman filtering, a risk-sensitive filtering is utilized for a linear time-varying fading channel estimation problem and its robustness against uncertainty parameter is explored. The time-varying channel estimation problem is formulated as time varying Finite Impulse Response (FIR) filter with known rectangular pulse at input and Gaussian noise corrupted signal at output. In time-varying FIR filter, estimation of time varying coefficients with uncertain conditions is an important and crucial task In literature, Kalman filter based fading channel estimation approach has been studied and it has limitations that leads to inaccurate estimation when there is high level of uncertainty in initial conditions and bias in system models and/or noise covariance. To overcome the above limitation, the risk sensitive filter is proposed. In this work only bias model is considered to investigate the effect of risk cost performance. The special feature of risk sensitive filter is, it uses a risk factor/parameter in the exponential cost function. So that the probability density function of variable is reshaped with proper tuning of risk factor, thus robustness again uncertainty can be achieved. In channel estimation with the uncertain conditions, the improved performance of risk sensitive filter over the conventional Kalman filter is shown with numerical simulation results.