The main problem of particle filter(PF)in nonlinear state estimation is the particle degeneracy.Resampling operation solves degeneracy to some extent,but it results in the problem of sample impoverishment.Variance red...The main problem of particle filter(PF)in nonlinear state estimation is the particle degeneracy.Resampling operation solves degeneracy to some extent,but it results in the problem of sample impoverishment.Variance reduction technique is proposed to deal with the degeneration phenomenon in this paper,which reduces the variance of the particle weights by selecting an exponential fading factor,and this factor can be chosen adaptively and iteratively in terms of the effective particle number.A theorem is presented to show that this idea is feasible,and the procedure of this new adaptive particle filtering(APF)algorithm is presented.Then,the principle of parameter choice and the limitation of APF are discussed.Finally,a numerical example illustrates that the proposed APF has a higher estimation precision than particle filter-sampling importance resampling(PF-SIR),genetic particle filter(GPF),and particle swarm optimization particle filter(PSOPF),while the computation load of APF is mild.展开更多
In order to solve particle degeneracy phenomenon and simultaneously avoid sample impoverishment, this paper proposed an improved particle filter based on fine resampling algorithm for general case, called as particle ...In order to solve particle degeneracy phenomenon and simultaneously avoid sample impoverishment, this paper proposed an improved particle filter based on fine resampling algorithm for general case, called as particle filter with fine resampling （PF-FR）. By introducing distance-comparing process and generating new particle based on optimized combination scheme, PF-FR filter performs better than generic sampling importance resampling particle filter （PF-SIR） both in terms of effectiveness and diversity of the particle system, hence, evidently improving estimation accuracy of the state in the nonlinear/non-Gaussian models. Simulations indicate that the proposed PF-FR algorithm can maintain the diversity of particles and thus achieve the same estimation accuracy with less number of particles. Consequently, PF-FR filter is a competitive choice in the applications of nonlinear state estimation.展开更多
基金Supported by National Natural Science Foundation of China(6063403060702066)Aerospace Science Foundation(20090853013)Doctoral Program Foundation of China(20060699032)
文摘The main problem of particle filter(PF)in nonlinear state estimation is the particle degeneracy.Resampling operation solves degeneracy to some extent,but it results in the problem of sample impoverishment.Variance reduction technique is proposed to deal with the degeneration phenomenon in this paper,which reduces the variance of the particle weights by selecting an exponential fading factor,and this factor can be chosen adaptively and iteratively in terms of the effective particle number.A theorem is presented to show that this idea is feasible,and the procedure of this new adaptive particle filtering(APF)algorithm is presented.Then,the principle of parameter choice and the limitation of APF are discussed.Finally,a numerical example illustrates that the proposed APF has a higher estimation precision than particle filter-sampling importance resampling(PF-SIR),genetic particle filter(GPF),and particle swarm optimization particle filter(PSOPF),while the computation load of APF is mild.
基金supported by the High-Tech Research and Development Program of China (2008AA7080304)
文摘In order to solve particle degeneracy phenomenon and simultaneously avoid sample impoverishment, this paper proposed an improved particle filter based on fine resampling algorithm for general case, called as particle filter with fine resampling （PF-FR）. By introducing distance-comparing process and generating new particle based on optimized combination scheme, PF-FR filter performs better than generic sampling importance resampling particle filter （PF-SIR） both in terms of effectiveness and diversity of the particle system, hence, evidently improving estimation accuracy of the state in the nonlinear/non-Gaussian models. Simulations indicate that the proposed PF-FR algorithm can maintain the diversity of particles and thus achieve the same estimation accuracy with less number of particles. Consequently, PF-FR filter is a competitive choice in the applications of nonlinear state estimation.
