| Being as a mathematical method, the model of state space can be used to describe the both of characteristics and states. Comparing the description based signal, the state space model is more popular with wide applications, such as information processing, identification of mode and automatic control. The estimation of state variable is one of key problems during modeling the state space model. And a series of algorithms have been presented. Especially, so-called particle filter without the restraints of linear and Gaussian has been developed quickly.Based on the analysis of theory and application for traditional particle filter, the jump Markov state space model is focused in this thesis. To estimating the states of two-dimension discrete and continuous in hybrid system, an approach with Rao-Blackwellised technology is proposed. By using the algorithm, some shortcomings in traditional particle filter have been overcome with accuracy improvement. The results of numerical simulation show that the algorithm presented here is effective and feasible. |
