Abstract

The particle filter offers a general numerical tool to approximate the posterior density function for the state in nonlinear and non-Gaussian filtering problems. While the particle filter is fairly easy to implement and tune, its main drawback is that its computational complexity increase quickly with the state dimension. By exploiting conditional dependencies between parts of the state to estimate, marginalized particle filter can improve the estimation quality while also reducing the overall computational load relative to standard PF. However, whether marginalized particle filter or other improving particle filters, process noise and observation noise are all supposed as the independent identical distribution. It will limit greatly the application field of sampling nonlinear filter. Aiming at the above problem, a decouple method of noise is structured by the rearrange and transformation of state transition equation and observation equation. Meanwhile, the decouple method is dynamically introduced into the framework of marginalized particle filter. Because the treatment for the computation-al complexity and correlation of noise are considered, the real-time and the filtering precision are obviously improved. Finally, the theoretical analysis and experimental results show the feasibility and efficiency of the proposed algorithm.

Paper Details