Abstract

Functional observers estimate selected linear combinations of system states, offering a computationally efficient alternative to full-state estimation. While their design has been extensively studied within the classical model-based framework, data-driven approaches remain comparatively underexplored, particularly in the presence of unknown inputs.

This talk will present a data-driven framework for designing unknown-input functional observers (UIFOs) for discrete-time linear time-invariant (LTI) systems. Beginning with a reformulation of the classical model-based design conditions, the talk will establish necessary and sufficient conditions for UIFO existence expressed entirely in terms of available input, output, and state data. Under suitable assumptions, the equivalence between these conditions and their model-based counterparts will be demonstrated.

In this talk, a systematic algorithm will be presented for computing the observer matrices directly from finite-horizon historical data, without explicit knowledge of the system matrices. It will also discuss how the framework extends to conventional unknown-input observers (UIOs). In addition, the presentation will highlight the potential of data-driven methods for developing estimation techniques that do not require explicit mathematical models of the underlying systems.

About the Speaker

Pankaj Vadhvani is a doctoral researcher in the Department of Mathematics at the Indian Institute of Technology Patna, India. He did an M.Sc. in Mathematics from Banaras Hindu University, India, in 2019. His research focuses on data-driven control, with particular emphasis on observer design and fault estimation schemes that do not require explicit knowledge of system model parameters.

Join the Zoom meeting: https://uni-kassel.zoom-x.de/j/61939676764?pwd=40oucVZKNkcf4xNdQaPwP2mWOnNXZt.1

Meeting ID: 619 3967 6764 Passcode: 212268