The Antiparticle Filter—An Adaptive Nonlinear Estimator

摘要: We introduce the antiparticle filter, AF, a new type of recursive Bayesian estimator that is unlike either extended Kalman Filter, EKF, unscented UKF or particle filter PF. show for classic problem robot localization AF can substantially outperform these other filters in some situations. The estimates posterior distribution as an auxiliary variable Gaussian which gives analytic formula using no random samples. It adaptively changes complexity uncertainty changes. equivalent to EKF when low while being able represent non-Gaussian distributions increases. computation time be much faster than same accuracy. have simulated comparisons two types iterative UKF, and PF demonstrating reduce error consistent accurate value.