Accuracy Analysis of Unscented Transformation of Several Sampling Strategies
Wu Zhang, Min Liu, Zhao Zong-gui · 2009
Unscented Transformation (UT) has been paid much attention within nonlinear filtering community for its significant accuracy and implementation advantages. Although several sampling strategies for the UT method has been developed, their comparison and analysis is still absent. Based on the concepts of multidimensional Taylor series expansion and moments, this paper studies unscented transformation accuracy of the sigma points sets generated from symmetric sampling, minimal skew simplex sampling and sphere simplex sampling strategies separately. The theoretical analysis shows that for Gaussian-distributed random vector whose components are independent of one another, the estimation accuracy of UT method based on symmetric sampling strategy is the highest while sphere simplex sampling strategy is the lowest. Interrelated simulations verified this conclusion.