Performance enhancement of the IMM estimation by smoothing

This paper proposes and investigates an enhancement of the IMM estimation (referred to as smoothing-enhanced IMM or SE-IMM) at the expense of a reasonable increase in computation. It is based on improved reinitialization of the IMM filter by IMM smoothing. An approximate IMM smoothing (IMMS) algorithm is provided. To maintain a minimum overall computational load, the SE-IMM scheme is applied only in the transitional regimes of the IMM filter when the dominant model changes. A cumulative sums (CUSUM)-type statistical test is formulated and employed for the problem of model change detection/estimation. An investigation of the efficiency of the proposed smoothing-enhanced IMM estimation is performed by simulation of a maneuvering target tracking scenario.