In order to solve the problem of performance degradation and divergence of the Gaussian nonlinear filter in the large initial deviation conditions,a new nonlinear filtering method called the iterated update extended Kalman filter (IU-EKF) was proposed.The new approach was carried out by introducing the current time measurement information gradually to the measurement update process with part of gain in pseudo-time. Meanwhile,since the multi-step measurement update process introduced the process noise at each step,the cross-covariance between the measurement noise and the posteriori state estimation error after each step was substituted into the covariance matrix,whose trace was then differentiated with respect to the standard Kalman gain and the result was set to zero.The optimal Kalman gain expression under correlated noise condition was derived then.At last,the number of iterations was adjusted adaptively according to the posteriori measurement residuals.In the premise of ensuring a certain filtering accuracy,the computational complexity of the algorithm was reduced.Taking the two-dimensional target tracking problem as an example,the algorithm was compared with EKF,IEKF,UKF and CKF respectively under the large initial deviation conditions.The influence of different iterations on the filtering accuracy was also compared and analyzed.The simulation results showed that the algorithm was more efficient than EKF,and the algorithm was superior to the classical Gaussian hypothesis filters under the conditions of large initial deviation.Furthermore,when the number of iterations was increased by 1,2,5,10,20,the filtering accuracy of the algorithm was improved,but the growth ratewas gradually slowed down.
