Interacting multiple models algorithm based on α-β & α-β-γ filters

In this paper, we firstly prove in interacting multiple models algorithm (IMM) model-conditional estimational variances can be decoupled with model-conditional state estimation under certain conditions, and then supply the conditions to partly-decoupling and absolutely-decoupling among model-conditional estimational variances. Therefore IMM based on α-β α-β-γ filters is proposed. The Monte Carlo simulations show that this algorithm not only remains almost as accurate as IMM, but also save about half of the computation.