A Research On Maneuvering Target Tracking Based On Interactive Multi-model Particle Filter Algorithm

With the development of science and technology,maneuvering target tracking plays an increasingly important role in modern military and civil fields.The improvement of target maneuvering performance and the increasingly complex tracking environment further promote the development of maneuvering target tracking technology.In the technology of maneuvering target tracking,the filter tracking algorithm is the key tool of maneuvering target tracking,so it is very important to study the filter algorithm deeply.Secondly,in practical applications,it is not enough to rely on single-sensor filtering algorithm for target tracking.Because the useful information of multi-sensor target tracking is more than that of single sensor,the final fusion estimation performance is better.Therefore,this thesis mainly studies the maneuvering target tracking technology from these two parts.The first part is the research on filter tracking algorithm.Considering the nonlinearity of stochastic dynamical system,the core idea of Particle Filter(PF)algorithm is to express the posterior probability density of state by weighted particles,which has no restriction on the state and noise of the system.Therefore,this thesis based on the interactive multi-model particle filter(IMMPF)algorithm from two different dimensions to improve.The first is from the perspective of integration of interaction layer and combination layer.Combining the definition of weighted KL divergence in information theory,an interactive multi model particle filter algorithm based on weighted KL divergence(KLIMMPF)is proposed by fitting the probability density function of the state estimate value.The probability distribution of the current target state can be approximated by the probability distribution of the state estimate value obtained from each model,and then the state estimate value of the target can be obtained.The second is from the perspective of feedback,since the fusion estimation of the final output often has a higher accuracy than the estimation dependent on each mode,it is considered as a reference term for state estimation at the next moment.An interactive multiple model particle filter algorithm with feedback learning item(FBIMMPF)is proposed.Finally,through theoretical analysis and simulation experiments,it is verified that the two improved algorithms from different perspectives are superior to the original IMMPF algorithm.In the second part,combined with the multi-sensor distributed fusion algorithm that is easy to implement in engineering,the IMMPF algorithm is introduced into the distributed multi-sensor network,and an interactive multi-model particle filter algorithm based on covariance intersection(CIIMMPF)fusion is proposed.Through theoretical proof and simulation experiments,it is verified that the multi model algorithm based on multi sensor fusion can improve the tracking performance of maneuvering targets.On the basis of the consistency of covariance intersection(CI)fusion estimation algorithm,an improved CI fusion estimation algorithm is obtained.From the analysis of simulation experiments,the IMMPF algorithm based on the improved CI fusion algorithm(improved CIIMMPF)also improves the tracking performance of maneuvering targets,and from the perspective of timeliness,the improved CI fusion algorithm is superior to the original CI fusion algorithm.