Robust detection of structural changes in nonlinear stochastic systems

The synthesis of a sequential structural-change-detection algorithm is presented with the emphasis on minimizing the priori information required. Such an algorithm is expected to be fairly robust with respect to a variety of probabilistic change models. The design is based on the class of conditionally linear signal generators that readily accommodate various feedback-control and parameter-estimation schemes. The proposed algorithm is optimal with respect to a wide class of sequential detection algorithms