Abstract
When tracking a maneuvering target with a three-dimensional radar, which can measure range, azimuth, and elevation of the targets, block processing managed by a Markov chain is often adopted under the framework of the interactive multiple model (IMM) estimator. However, tracking performance may degrade because different maneuvers influence each other among three Cartesian coordinates. In order to solve this problem, this paper presents a separated maneuvering target tracking algorithm. First, it formulates the diagonalization and minimization of the upper bound of the modified converted measurement noise covariance matrix as an optimization problem. Then, the existence and uniqueness of the solution for this optimization problem are demonstrated. Finally, it exploits the IMM algorithm to estimate the target states along with each Cartesian coordinate separately. The tracking performance of the proposed algorithm is compared with the square root IMM with L-D factorization, IMM with sequential importance sampling particles, and multiple-model estimation with model-group switching. Monte-Carlo simulation results demonstrate that the root mean square error of the proposed tracking algorithm is lower than that of the above algorithms for maneuvering target tracking.