A new class of particle filters for random dynamic systems with unknown statistics (Q1773797)

The goal of the paper is the new methods of estimation of an unobserved discrete-time random signal in a dynamic system. The authors introduce a new particle filtering (PF) approach to deal with uncertainty in the probabilistic modeling of their random dynamic system. Their approach does not need explicit mathematical forms of the probability distribution of the noise in the system. The methods are robust and flexible. A new family of PF algorithms is proposed, in which the statistical reference of the a posteriori state probability density function is substituted by a user-defined cost function that measures the quality of the state signal estimates according to the available observations. The problem of identifying sufficient conditions of asymptotically optimal propagation of particles is discussed, the same as the problem of local resampling. Computer simulation results are presented.