In particle filter, the use of resampling technique solves the problem of particle degeneracy to some extent, but introduces a new problem of particle impoverishment. To reconcile this dilemma, a dynamic resampling strategy is proposed, where the resampling operation is only performed on part of the particles in a step-by-step manner and the number of resampled particles is determined dynamically by a termination criterion based on the effective sample size. The new-born particles produced by resampling operations are helpful in alleviating particle degeneracy, whereas particles free from being resampled is conducive to improving the diversity of particles. Thus, a tradeoff between these two problems can be achieved. Also, two techniques, that is, recursive computation of the effective sample size and improved bisection method, for improving the computational efficiency are proposed. Simulation results conducted on two typical examples show the improved performance of the proposed method.
[1]
Tiancheng Li,et al.
Deterministic resampling: Unbiased sampling to avoid sample impoverishment in particle filters
,
2012,
Signal Process..
[2]
Monish D. Tandale,et al.
Particle Filter for Ballistic Target Tracking with Glint Noise
,
2010
.
[3]
Yaakov Oshman,et al.
Adaptive Particle Filtering for Spacecraft Attitude Estimation from Vector Observations
,
2009
.
[4]
Neil J. Gordon,et al.
A tutorial on particle filters for online nonlinear/non-Gaussian Bayesian tracking
,
2002,
IEEE Trans. Signal Process..
[5]
M. Pitt,et al.
Filtering via Simulation: Auxiliary Particle Filters
,
1999
.