Publications (1)0 Total impact
ABSTRACT: This paper describes the scaled unscented transformation, a new method of applying the unscented transform to a nonlinear system. A set of samples are deterministically chosen which match the mean and covariance of a (not necessarily Gaussian) probability distribution. Each point in the set is scaled by a user-specified constant. A method is derived which preserves the second order accuracy in mean and covariance, giving performance as good as second order truncated filter. 1 Introduction One of the most fundamental tasks in filtering and estimation is to calculate the statistics of a random variable that has undergone a transformation. When the system models are non-linear no general closed-form solutions exist  and many approximations have been proposed [1--3, 12, 13]. In  and  we introduced a new approximate method for propagating means and covariances through nonlinear transformations called the unscented transformation. A set of weighted sigma points are deterministical...