The kalman filter described above predicts a random process state variable Xt∈Rn described by a linear random difference equation, then what should be done if the system model is nonlinear?The Extended Kalman Filter gives a solution to this situation. In the face of nonlinear relations, we linearize by finding the partial derivatives of the process equation and the measurement equation, and calculate the current predicted value.Different from the basic Kalman filtering (KF) process, the factor matrix (A, W, H, K) in the extended Kalman filtering (EKF) process varies from time to time.The basic working steps of the extended filter (EKF) are the same as those of the basic filter. The main difference between the two is that linearization is needed in the non-linear case, and the factor matrix generally changes with time (related to time T).