Abstract: A new two-dimensional target maneuver model is presented. Tangential and normal accelerations are in the state vector. An analytical solution is presented to compute the state transfer function when both normal and tangential acceleration are constant. This model can give a better priori acceleration estimation. Tangential and normal accelerations are driven by white noise respectively in the model, and have different noise levels. The new filter algorithm provides a better performance, especially in the acceleration estimation.

Key words: maneuvering target tracking, Kalman filter, two-dimensional horizontal motion model, coupled model, state estimation