Improved Particle Filter Algorithm And Its Application In Maneuvering Target Tracking

As we all know,the maneuvering target tracking technologies have made in-depth research and remarkable development at home and abroad as widely usage in the military and civilian and other fields.Kalman filter and Extended Kalman filter are the most typical filter algorithms in maneuvering target tracking domain,of which the former under linear and Gaussian system can achieve the optimal state estimation in the Minimum-Mean-Square-Error sense,and the latter under the weak nonlinear system can obtain suboptimal estimation.Kalman filter and Extended Kalman filter estimation precision is hard to guarantee,or even lead to filter divergence when strong nonlinear,non-Gaussian distribution occurs.Particle filter,which is developing rapidly in recent years,as a kind of nonlinear filtering algorithm,and being able to handle arbitrary state estimation problem of nonlinear,non-Gaussian system have been under the spotlight.The paper mainly studies the principle of particle filter and other typical nonlinear filtering algorithm,and its applications to maneuvering target tracking,in order to effectively realize the maneuvering target tracking.Firstly,this paper details the basic principles of a variety of nonlinear filtering algorithm,analyzes their advantages and disadvantages and the scope of application.Secondly,to solve the problem of particle degradation,a new auxiliary particle filter algorithm is proposed.New method obtains the sampled particles from the successive approximation of the posterior probability distribution by introducing the regularization step into the resampling process,which increases the diversity of particles and improve the estimation precision.An improved Gaussian particle filter method based on the cubature kalman filter is proposed.The importance density function is structured by using cubature kalman filter.Including the latest observing information.In the time updating stage,the CKF algorithm is used to update the Gaussian distribution parameters only in order to improve the efficiency of the algorithm.The MATLAB simulation results show that the proposed method feasibility and validity.Finally,aiming at the maneuvering target tracking problem,this paper summarizes the basic principle of maneuvering target tracking technology,introduces some commonly used target motion model and interacting multiple model,because of Interacting Multiple Model can accurate and effective identification of motor parameters,and through multiple model to describe the motion model in maneuvering target tracking.And proposed method is compared with the Interacting Multiple Model to solve the maneuvering target tracking problem under different intensity.The simulation results show that IMM-CKGPF can effectively realize the target tracking under different maneuvering conditions,and it has better tracking performance than IMM-GPF.