Two stage prediction and update particle filter with correlated noise in multi-sensor observation

Particle filter realizes recursive Bayesian filter via Monte Carlo simulation. The method is suitable for any non-linear system that could be represented with state model. However, the precision of particle filter depends mainly on two key factors, the effective sampling of particles state and the reasonable measuring of particles weight. In addition, considering the correlated noise also occurs in practical application, the basic assumption of particle filter sometimes can not be met, which affects on directly stability and reliability of filtering result. Aiming at the above problem, a novel two stage prediction and update particle filtering algorithm with correlated noise in multi-sensor observation is proposed in this paper. Firstly, in order to avoid adverse influence from the correlation between observation and process noise for filtering precision, the system model is modified by rearrange the state transition equation and the observation equation. Secondly, considering the rational utilization of multi-sensor observations, the weight fusion strategy of particle weight is used to weaken the adverse influence from random observation noise in measuring process of particle weight, and the two stage prediction and update framework is constructed to realize the optimization of sampling particle state by the introduction of latest observation. Finally, the theoretical analysis and experimental results show the feasibility and efficiency of proposed algorithm.