Unscented particle filter using scaled spherical simplex UKF

In order to reduce the computation burden of conventional unscented particle filter, a method for particle filter based on spherical simplex unscented transformation (SSUT) is proposed. This method uses spherical simplex unscented Kalman filter to generate importance distribution of particle filter. It can extend its overlaps and posterior probability density, and reduce the computation burden by reducing sigma points. However, the sigma point set coverage radius expends over dimension of state space, which results in the deterioration of the aggregation of sigma points. Auxiliary random variable formulation of the scaled transformation can overcome the defect of sigma point set distribution expansion. So the scaled spherical simplex unscented particle filter (SSSUPF) is introduced. The simulation results show that compared with conventional unscented particle filter (UPF), the computation complexity of SSSUPF can be reduced by 50 percent, and compared with spherical simplex unscented particle filter (SSUPF), SSSUPF reduces the system noise and the measurement noise variance estimation error.