2020/04/02 by Mohammad Khajenejad, Khajenejad, Mohammad, Zeyuan Jin +3
Biochemistry, Genetics and Molecular Biology · Engineering · #Advanced Control Systems Optimization #Control Systems and Identification #FOS: Electrical engineering #Receptor Mechanisms and Signaling #Systems and Control (eess.SY) #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2004.03665
openalex publication_date 2020/04/02 · openalex created_date 2022/07/26 · openalex updated_date 2026/07/28
We study the problem of designing interval-valued observers that\nsimultaneously estimate the system state and learn an unknown dynamic model for\npartially unknown nonlinear systems with dynamic unknown inputs and bounded\nnoise signals. Leveraging affine abstraction methods and the existence of\nnonlinear decomposition functions, as well as applying our previously developed\ndata-driven function over-approximation/abstraction approach to over-estimate\nthe unknown dynamic model, our proposed observer recursively computes the\nmaximal and minimal elements of the estimate intervals that are proven to\ncontain the true augmented states. Then, using observed output/measurement\nsignals, the observer iteratively shrinks the intervals by eliminating\nestimates that are not compatible with the measurements. Finally, given new\ninterval estimates, the observer updates the over-approximation of the unknown\nmodel dynamics. Moreover, we provide sufficient conditions for uniform\nboundedness of the sequence of estimate interval widths, i.e., stability of the\ndesigned observer, in the form of tractable (mixed-)integer programs with\nfinitely countable feasible sets.\n