A Fast and Unbiased Minimalistic Resampling Approach for the Particle Filter

Abstract

The particle filter is an important approximation method for online state estimation in nonlinear nonGaussian scenarios. The resampling step in the particle filter is critical because it eliminates the wasteful use of particles that do not contribute to the posterior (degeneracy). The fully stochastic resamplers, despite being unbiased in approximating the posterior density, involve exhaustive and sequential communication within the particles and thus are computationally expensive. The alternate partial deterministic resamplers overcome this problem by reducing the communication within particles but this leads to approximation bias. This paper proposes a fast resampling procedure that gives an accurate approximation of the posterior and tracks as accurately as the conventional resamplers.