Study of State Estimation with Super Particle Filter

Model with nonlinear and time variance is often met in the question of time sequence state estimation. It is not very good to solve it by EKF algorithm. We propose a new algorithm, super particle filter (SPF), which adds hyper parameters in state vector and estimates state and parameters simultaneously online. By the method, hype parameters can be adjusted to change with model automatically. The introduction of hyper parameters to state vector makes state space model nonlinear. For the reason, particle filter is applied to solve the nonlinear and non-Gaussian state space models. We compared this algorithm to the EKF algorithm. Experimental results show SPF algorithm increase 60% in accurate and 70% in time expenditure