To improve tracking accuracy of maneuvering target

an i nteracting multiple model self-adaptive unscented particle filter algorithm bas ed on predicted residual was proposed to deal with the nonlinear and non-Gaussi an systems.The alorithm combined interacting multiple model with improved parti cle filter

where unscented transformation was used to generate a proposal distr ibution function.The measure-covariance and state-covariance were automatic a djusted by predicted residual online

thus improving estimation accuracy and red ucing the calculation error.Simulation results reveal better performance tracki ng accuracy of the proposed algorithm in comparison with traditional interacting multiple mode particle filter and unscented particle filter.