On disorder detection in linear stochastic system via noisy observations

The problem of change point detection for multidimensional stochastic discrete-time process is considered. The process is described by stochastic difference equation with Gaussian noises. The parameters of process are changed abruptly at an unknown point in time. The sequential disorder detection procedure is proposed. the analysis of this procedure is carried out. The formulae for mean time between false alarms and for mean delay in disorder detection are derived. Asymptotical relation between these characteristics is established. Proposed procedure does not require the accumulation of data and, hence, is very simple for practical realization.