Fast and effective generation of the proposal distribution for particle filters

The use of particle filters to solve non-Gaussian, nonlinear estimation problems has attracted the attention of many researchers in recent years. The particle filters require a proposal distribution, and formulation of the proposal distribution is a critical design issue. Here a fast and effective method for generating the proposal distribution is derived on the basis of the extended H∞H∞ filter. The resulting particle filter, called the extended H∞H∞ particle filter, provides performance comparable to that of the unscented Kalman particle filter but with lower computational cost.