Online modeling refinement for discrete event systems

Machine identification of discrete event systems (DES) addresses the issue of identifying an unknown system based on externally observed sample path of the unknown system. Online Modeling Refinement studies the continuing machine identification process in the context when the observed sample path is updated incrementally. While machine identification problem for fixed length sample path is NP-complete, the computational requirement for the proposed online modeling refinement algorithm is maintained at minimal by taking the structure similarity between successive accumulated observed sample paths. In addition to the computational advantage, the proposed algorithm also guarantees the identification results of the system models "converge" to the unknown DEDS model as the incrementally observed sequence get "long" and "rich" enough.