Input-Adaptive Models Based Multiple-Model Algorithm for Maneuvering Target Tracking

The dynamic models with multilevel inputs are adopted in a kind of multiple model estimator for highly maneuvering target tracking. While the target maneuvers with the continuous time-varying accelerations, the estimator increases the levels to improve the percentage of coverage, which induces two problems: the increase of calculation burden and the decrease of the estimation precision due to the competition between the models. A multilevel input-adaptive multiple model (IAMM) algorithm is proposed, in which the inputs are adjusted according to the prior value and the on-line estimated maneuver parameters by introducing a dualistic distribution. The adaptabilities of the inputs can depict the actual maneuver process better compared with the static multilevel inputs. The simulation proves the effectiveness of IAMM algorithm compared with the IMM (Interacting Multiple Model) estimator with models containing multilevel static inputs.