Abstract:

Aiming at the effective utilization of multi-sensor observation and the simplification of computational complexity for the state estimation of nonlinear systems in observation uncertainty, a novel multi-sensor observation adaptive Rao-Blackwellised particle filtering algorithm is proposed. Firstly, in the new algorithm, the effective sensor observation set used to measure particle weight is sampled by means of the random sampling strategy and the observation model prior transition probability. Then the identification of the sensor observation model without the influence of disturbance is realized on the basis of the re-sampling steps and the probability maximization principle. Finally, according to the independently solving way of nonlinear state component and linear state component in Rao-Blackwellised particle filter, the system state estimation is achieved at current time. The theoretical analysis and experimental results show the feasibility and efficiency of the proposed algorithm.