Research And Application Of Interacting Multiple Model Algorithm With Adaptive Transition Probability Matrix

Interacting Multiple Model(IMM)is a mainstream and the most cost-effective multiple model algorithm with low computational complexity and high estimation accuracy.It has a wide range of applications in the field of hybrid system state estimation.Transition Probability Matrix(TPM)is an important parameter in IMM algorithm,which is often set to a constant matrix according to the prior information.This is a conservative parameter setting method which makes IMM algorithm fail to achieve the desired performance.Aiming at the research and application of IMM algorithm based on adaptive transition probability,the following work has been carried out:1.Because the model jump trend is contained in the slope of the model probabilities,a transition probability correction function is constructed according to this slope.An adaptive IMM algorithm based on slope of model probability is proposed.The simulation experiments of maneuvering target tracking are conducted to verify the performance of proposed algorithm.2.According to the latter difference theory,a new transition probability correction function is constructed,and adaptive IMM algorithm based on N-order latter difference of model probability is proposed,the influence of differential order N on the performance of the algorithm is discussed.The simulation experiments of maneuvering target tracking are conducted.3.Aiming at the problem that proposed adaptive IMM algorithm modified by the model probability information can not judge current model state of hybrid system and leads to the decrease of response ability to models jump,a model jump judgment method based on likelihood function ratio is proposed,and the transition probability correction function is reconstructed by this method,and then an adaptive transition probability matrix based parallel IMM algorithm is proposed.Model switching threshold is an important parameter for this algorithm,its setting guidance is given through theoretical analysis and simulation experiments.4.The motion state analysis of train is actually a kind of state estimation of hybrid system.Aiming at the difficulty of determining the parameters of dynamic model set,a dynamic model set modeling method of train based on force analysis is proposed,and then the adaptive transition probability based IMM algorithm proposed in this thesis is applied to estimate the motion state of train.Simulation results show the effectiveness and feasibility of the modeling method and adaptive transition probability based IMM algorithm in the motion state estimation of running train.