Sigma particle filtering for nonlinear dynamic systems

In this paper an efficient particle filtering algorithm is derived for nonlinear estimation. The new filter is called the sigma particle filter and is formulated by combining a sigma particle sampling method with the sequential weight update from particle filtering. It draws sigma particles deterministically from various sigma boundaries instead of taking random samples from a selected probability distribution. The sequential weight updates are carried out by utilizing the measurement likelihood function used in the particle filter. The new sigma particle filter has advantages over the standard particle filters in that it not only mitigates the computational load, but also provides results as accurate as those obtained by the standard particle filters. The performance of the new particle filter is demonstrated through a reentering spacecraft example.