Evaluation of the log-homotopy particle flow filter for localisation

Nonlinear state estimation related to systems with high dimensional state spaces is a difficult problem that has theoretical and practical relevance. The performance of the Kalman filter is consistent in any dimension, however the same does not hold for the particle filters, due to the curse of dimensionality and particle degeneracy. The log-homotopy based particle flow filter, introduced in recent years is a promising approach to hard state estimation problems. This paper examines the applicability of the Daum–Huang particle flow filter in a challenging multi-target tracking scenario involving a nonlinear measurement model and compares it with conventional estimation methods.