Target Tracking In Nonlinear Observer Under The Conditions Of Its Decoupling Study

The target is measured usually under the polar or spherical coordinate in the target tracking system, while the state equations are given under the Cartesian coordinate. This paper studies the target tracking subjects which is under those non-linear measurement conditions, and gets a coupled filter. To reduce the workload of coupled filter, make it independent in each direction, the paper studies and gets it's decoupled algorithm which can analyse the filter's performance more conveniently.Firstly, the paper introduces in brief the Bar-Shlom's CMKF-D algorithm which is studied under the polar coordinate. Then the paper extends the algorithm to the spherical coordinate, and computes the mean and convariance of the measurement error after the coordiate has been changed to be the Cartesian.With the convariance of the measurement error, the paper brings forward a new method on how to get the initial estimations of target tracking under above-mentioned condition. The simulations prove that the method has a good result no matter the measurement error is great or small.The CMKF-D filter is coupled for its measurement noise of every direction is correlative, and its measurement noise covariance is non-diagonal. The paper firstly introduces a canonical transform which can diagonalize the process noise and the measurement noise covariance simultaneously. Illuminated by this idea, the paper puts forward a decoupled CMKF-D alogrithm based on the canonical transform, whose transform matrix is computed in alalytical expression. We can analyse the filter conveniently through the decoupled CMKF-D alogrithm for the filter is independent in all direction. The simulations prove that the decoupled alogrithm still has the same good performance with the original one.