LAUSR

Particle filters are vital tools to solve the nonlinear/non-Gaussian filtering problems. However, it suffers from the problems of sample degeneracy and impoverishment. This paper proposes a framework of particle filter with a hybrid sampling strategy, generates particles by means of deterministic and random sampling, and then, restores the fixed number particles by a resampling method. Compared with traditional sampling strategies in the particle filters, the hybrid sampling framework is more suitable for the characteristics of a time-varying system. In order to demonstrate the effectiveness of the proposed framework, a novel particle filter based on hybrid deterministic and random sampling (HDRSPF) is designed, where the deterministic sampling strategy is UT transform and the random sampling strategy is Mont Carlo sampling. Simulation results of single-dimensional and multi-dimensional scenarios show that the proposed filter HDRSPF has a higher filtering accuracy at the cost of lower computational complexity as compared with the existing methods.