Properties of risk-sensitive filters/estimators

Algorithms for risk-sensitive filters have been developed in literature and connections to H∞ filtering also established. The risk-sensitive filter differs from a conditional mean estimator (Kalman filter) and is either risk-prone or risk-averse depending on the sign of a scalar thetas that appears in the cost function. The RS filter exhibits many interesting properties. Statistical properties, parameter estimation and explicit bounds of estimation for these filters are presented in the paper.