Research on particle filter based on spherical unscented transformation

In order to improve the particle degeneracy phenomenon of particle filter, a method for particle filtering based on unscented transformation was proposed. The spherical unscented Kalman filter was used to generate the important distribution for particle filter. The important distribution integrated the latest observation, so it can extend the overlaps of itself and posterior probability density and well approximate the true distribution of the state. The spherical unscented Kalman filter had same accuracy as generic unscented filter but required nearly half samples. The simulations results show that compared against widely used unscented particle filter (UPF), the computation of new algorithm can be reduced by 50 percent and the computation time can be reduced by 34 percent. So the new algorithm was an effective nonlinear estimation method.