The Algorithm Research Of Maneuvering Target Tracking Based On Interacting Multiple Models

Target tracking is a technology that estimate and forecast the target’s current and futurestate according to the target information obtain by various sensors. And target tracking iswidely used in military and civil area. With the advancement of science and technology,various new technology and theory was applied to target tracking And Target trackingtechnology gradually developed into an interdisciplinary, cross-sectoral, multi-leveltechnology.The main study content of target tracking is restoring accurate information frominaccurate information, so a filter is needed to process the data. Because of the excellentperformance, Kalman filter was widely used in target tracking domain. And the filter is basedon certain mathematical models. So the mathematical model and the filter algorithm is the twomain contents of target tracking. Many experts and scholars home and abroad conducteddeeply research on it, and got many achievements.As a new data fusion algorithm, Interaction multiple model (IMM) algorithm has receivedenough attention because of its excellent tracking performance and wide filtering bandwidth.But, most researches mainly focused on the improvement of the model interacting, data fusion.With the continuous development and maturity of fuzz theory and neural networks, these newtheories are also applied to the IMM algorithm, and greatly promoted the development ofIMM algorithm. This paper is also conduct the research based on IMM algorithm.Firstly, This paper introduces the basic principles of target tarcking,and some basicfiltering algorithms.And mainly studied the least squares estimation, filter,Kalmanfilter,extended Kalman filter,whilc is extensively used in filtering theory.Through thesimulation of tracking a constant velocity target,the tracking results of filter andkalman filter was tested.Secondly, some typical Mathematical model was introduced into this paper. such asCV,CA models which applied to tracking Non-maneuvering target, and Singer model,“currentstatistic” model, Jerk model, which applied to tracking maneuvering target. At the same time,as the main study content, this paper analyzed the fundamentals of the IMM algorithm.Because of the increment of filtering bandwidth rely on more models added into the model set.But too many models will not only increase the computation of the algorithm but also reduce the accuracy of the algorithm for the competition of different model. To solve this problem,this paper started the study from the following aspect.(1)Because of the maneuvering of the target largely reflected in the fluctuation of the targetacceleration. so this paper mainly focus on improving the estimate precision of acceleration.Taking the acceleration estimating result of “current statistic” model into account, this paperdesign a new algorithm which introduce Amax adaptive to “current statistic” model toimprove the performance of the acceleration estimation.(2)To get the wide bandwidth of IMM algorithm, this paper designed a fuzzy inference basedalgorithm, which refered to variable structure multiple model algorithm. The algorithm usethe improved “current statistic” model to conduct the pre-estimate of acceleration. Thenaccording to the estimating result of acceleration, Appropriate model was added to the modelset, Those model which didn’t meet the motion mode will be excluded to the model set, so thealgorithm will got a better balance between Real-Time and the filtering bandwidth.(3)To test the result of improved algorithm, a simulation between standard IMM algorithmand the improved algorithm was carried out. And the result show the improved algorithmreached the desired objective.