Research On Identification Algorithm Of System With Multiplicative Parameter Based On EM

Maneuvering target tracking is the precondition of precision strike in military field and air traffic control in civil field and the turn rate(TR)identification is the key of Maneuvering target accurate tracking.Existing methods can only approximately identify it and the identification accuracy is poor,owing to the nonlinearly couple characteristic of turn rate.Therefore,in order to improve the identification of turn rate,this thesis modeled the issue of turn rate identification in maneuvering target tracking as the one of multiplicative parameter identification of a class of dynamic system and discussed the analytical implementation and efficient implementation of Expectation Maximization(EM).The main contributions are as follows:1.As the poor identification accuracy of existing methods in TR identification,this thesis modeled maneuvering target tracking as a class of linear dynamic system with multiplicative parameter and proposed TR analytical identification algorithm.By linear equivalent transformation,TR nonlinearly coupled in dynamic transition matrix is equivalently transformed into a class of simple multiplicative parameter.Then,we applied EM algorithm to analytically identify the parameter and deduced the analytical expression of turn rate by linear equivalent inverse transformation.Simulation analysis demonstrated the efficiency of our algorithm.2.Based on the TR identification in maneuvering target tracking,this thesis further discussed joint state estimation and multiplicative parameter identification of a more complicated nonlinear system with the colored measurement noise.An adaptive divided difference filter was proposed to solve the problem of the correlation factor identification of colored measurement noise.In order to cope with the time varying characteristic of the TR and the correlation factor,we proposed a measurement truncated EM algorithm,in which the measurement in a certain window length from current moment to previous moment are selected to redesign the framework of EM algorithm,rather than all the measurement.More precise divided difference filter were applied to achieve fast and highprecision identification of time-varying multiplicative parameter.Simulation analysis demonstrated the efficiency of our algorithm.3.As EM algorithm is weak in implementation,this thesis discussed the design of adaptive window-varying(AWV)based on the detection of measurement variations and event-triggered measurement schedule.Two kinds of AWV scheme were proposed to solve the problem.Based on a novel stochastic event-triggered measurement schedule,we designed an adaptive window-varying scheme.Then,by embedding the AWV scheme into the framework of EM algorithm,we designed an AWV-EM algorithm based on event-triggered measurement schedule.The AWV-EM algorithm based on eventtriggered measurement schedule can improve the implementation efficiency of EM algorithm and reduce calculation burden with a little of accuracy loss.Simulation analysis demonstrated the efficiency of our algorithm.