Research on Unknown Static Parameter Estimation Problem of State Space Models Based on Particle Filter Algorithm

摘要

A novel parameter estimation method for unknown static parameters of the state space model using particle filtering (PF) has proposed in this paper. Traditional methods enlarge state vector by treating the unknown parameter θ as a part of state vector (,) k x θ. But this may cause the degeneration of θ, when some estimates become too small to continue as a result of the non-dynamic character of parameters if θ at time k is only determined by time k ?1. Compared to traditional methods, this novel method assumes that the posterior distribution of θ is given by previous observation and state vectors, Z_(1:k) and X_(1:k). Obtain statistics at time k by using the integration of Z_(1:k) and X_(1:k), and solve parameter estimation problem by updating θ recursively. Good results are obtained when this method is used in different models.