Research On Multi-model Distributed Maneuvering Target Tracking

Maneuvering target tracking is an important issue in military,navigation and transportation.On the one hand,a growing number of highly maneuverable targets appare on the modern battlefield.Therefore,it's importent to accurately track the highly maneuverable targets.On the other hand,information fusion has become a hot issue with the increasing number and types of sensors being used.Distributed information fusion architecture has attracted more and more attention because of its unique advantages such as non-convergence center,low communication bandwidth requirement and strong robustness,which is widely used in wireless sensor networks and other fields.Based on the above background,this paper is devoted to the research of maneuvering target tracking based on multi-model algorithm,and try to introduce the multi-model based maneuvering target tracking algorithm into a distributed sensor network through consensus-based filter.The main works are summarized as follows:1.Introduced the theory of hybrid systems and multi-model algorithm.To deal with the nonlinearity of measurement models of radar and infrared sensor,the unscented Kalman filter(UKF)is introduced.It is found that there is singularity in measurement functions of radar and infrared sensor,which will cause a sharp deterioration of state estimation.Therefore,the Azimuth-Singularity Free Kalman Filters are proposed.2.Two kinds of variable structure multi-model(VSMM)algorithms are studied in detail: Expected-Mode Augmentation(EMA)algorithm and Likely-Model set(LMS)algorithm.Their advantages and limitations are analyzed through simulation.Aiming at the insufficiency of the dynamic error of the LMS algorithm in the case of mode mutation,the improved LMS is proposed.The Expected-Mode Augmentation based on Likely model-set(EMA-LMS)is proposed.EMA-LMS can not only improve the estimation effect of the EMA algorithm,but also reduce the number of models used in each algorithm cycle.3.For the distributed radar and infrared sensor networks,the consensusbased Unscented Information Filter(DUIF)is introduced.The state estimation effect of three kind of DUIF algorithms are compared.In addition,considering the multi-model algorithms are better algorithms for maneuvering target,three kind of distributed interactive multi model(DIMM)are compared.Finally,distributed variable structure multi-model(DVSMM)is proposed to improve the performance of maneuvering target tracking in distributed networks.