High Order Multiple Model State Estimating Algorithm And Its Application

In the field of maneuvering target tracking,due to unable to get what the real mode situation of the maneuvering target at every moment is and when the mode changes to another,model uncertainty is a key problem for maneuvering target tracking with high precision.Currently,multiple model algorithm for estimating the maneuvering target state is a prevalent and effective way to deal with this problem,Among that the interactive multiple model filter is a most classical approach.In order to utilize more prior information to improve the estimating performance,this paper studies the high order multiple model algorithm,mainly making research on the following three aspects in detail.Firstly,studying the high order interacting multiple model filter based on mixture transition distribution algorithm.Compared with interacting multiple model algorithm,high order multiple model filtering algorithm improves the estimation accuracy by using more continuous multiple time information.But high order Markov chain needed in that has an obvious shortcoming that the number of independent parameters herein is too large,and it does not have sufficient prior knowledge to determine the high order model transition probability matrix.This paper adopts the mixture transition distribution model.It uses empirical weighted sum of the first order Markov model transition probability to approach high order model sequence transition probability and greatly reduces the number of required parameters,which reduces the difficulty of the determining high order model transition probability matrix reasonably.Simulation results show the effectiveness of this algorithm.Besides,the higher of the order is,the better of the filtering performance during steady area will be.Secondly,proposing a switching constrained high order multiple model filtering algorithm.High order Markov chain implies an assumption that model switching may occur at any time instance,which is not reasonable in actual scenarios since target does not usually change its mode at all times.So,based on high order Markov chain,this paper increases a constraint that the number of model switches is limited,that means there is no more than one model switching over a continuous period of time.Therefore,it gives a more accurate description of the model sequence,and improves estimation accuracy.At the same time,the computational efficiency of the algorithm is improved due to the deletion of many model sequences that do not meet the requirements under the constraint.Besides,cooperating with switching constrained high order multiple model filter algorithm,this paper also designs a more reasonable high order model sequence transition probability.The high order generalized pseudo Bayesian algorithm and high order interactive multiple model algorithm with model switching constraints are derived.The simulation results show that the estimation accuracy of this algorithm during model invariant region is extremely close to that of the algorithm known for model sequence,only in the model switching points existing a peak error;compared with the interacting multiple model algorithm,the estimation accuracy in all areas are improved;compared with high order multiple model filtering algorithm in same order,The error during model jump region is greatly reduced,the transition process is shortened greatly,and a large amount of calculation is saved.Lastly,proposing a switching constrained augmented state high order interacting multiple model smoothing algorithm.Based on the switching constrained high order multiple model filtering algorithm,this algorithm uses augmented state,performs smoothing while filtering.Simulation results show that compared with the augmented state interacting multiple model smoothing algorithm,smoothing effect of this algorithm can be further improved;compared with switching constrained high order multiple model filtering algorithm,it has better estimation accuracy,and basically eliminate peak errors.Besides,by setting different length of fixed lag,it has different smoothing effect.