DNKF 2013 - Special Session on Deterministic Sampling for Nonlinear Kalman Filtering

Non-deterministic sampling methods for nonlinear state estimation, such as Particle Filters (PFs) or Sequential Monte Carlo (SMC) methods, have gained popularity in the last several years. Nevertheless, their computational complexity, attributed to the exploited law of large numbers, is still a major problem. In consequence, they are not well suited for real time applications or high dimensional problems due to insufficient state space coverage. In order to overcome this drawback, a common and widely used approach is to choose samples in a deterministic manner by minimizing a certain distance measure, e.g., approximating moments of a Gaussian distribution. Popular examples include the Unscented Kalman Filter (UKF) and its derivatives, e.g., employing adaptive sampling. This session encourages all researchers to share new or optimized sampling approaches and their usage in the field of nonlinear state estimation.