Unscented Information Filtering with Interacting Multiple Model for Multiple Sensor Target Tracking

Abstract

A new method that combines an unscented information filtering (UIF) algorithm with an interacting multiple model (IMM) framework under a distributed. multiple-sensor fusion architecture is proposed. The objective of the proposed scheme is to track a maneuverable target whose dynamics can be modeled with multiple nonlinear models, and whose measurements are obtained from and processed at distributed systems. An IMM is not suited for information fusion architectures because it does not use combined estimates and covariance from a previous step to predict values at the next time step, which is essential for information filtering. The proposed algorithm fuses data, such as the information state contribution and information matrix, of each UIF that is included in an IMM filter. Moreover, the proposed algorithm improves the tracking performance when the mode likelihood functions in the IMM, which are important in flight mode detection and change, are shared among the distributed systems. The tracking results from simulations indicate that the present filtering method can be a good solution to tracking of a maneuvering target in multiple-sensor environments.