Searcharxiv⌕ Search

arXiv subjects

Stefan Damjanovic

Publications and source records attributed to Stefan Damjanovic.

2 recordsLinked to original sources

Dual-Mode Extremum Seeking for a Class of Unknown Controlled Diffusions

A dual-mode extremum-seeking controller is developed for unknown nonlinear systems with affine control inputs driven by persistent Brownian process noise. The controller uses a single measured cost, without state estimation or differentiation of the measurement. Bounded cosine feedback generates, through Lie-bracket averaging, a proportional--integral (PI) action: the proportional action stabilizes a feasible minimum of the cost on the time scale of the plant, and the integral action adapts a bias toward the unknown input that regulates the plant at that minimum. The averaged closed loop is a PI-regulated diffusion that retains the process noise, and its stability condition does not require knowledge of the control direction. Under a regional Lyapunov condition on the averaged dynamics, an exact periodic response of the noise-free closed loop is constructed and serves as the center of a Lyapunov function with bounds uniform in the excitation frequency. Exit-probability estimates show that, for small noise and large excitation frequency, trajectories reach a prescribed neighborhood of the optimum and remain there for times exponential in the inverse noise variance, with exponentially small failure probability. Under an additional stochastic contraction condition, the closed loop is approximated uniformly in time, in mean square before common exit, by its averaged diffusion driven by the same Brownian motion, with an error that decreases with the excitation frequency. The regional regulation conditions are verified for a planar hydrodynamic trap with an unknown stagnation-point offset and actuator sign. Simulations illustrate intensity-only confinement of a Brownian particle, integral compensation of the holding input, and axial-strip exit statistics.

math.OC↗

Distributionally Robust Frequency Response Estimation Using Non-Steady-State Frequency-Domain Data

This paper develops a distributionally robust optimization (DRO) framework for identifying the frequency response function (FRF) of an unknown linear time-invariant system from repeated noisy non-steady-state input-output measurements. Building on a recent frequency-domain extension of Willems' fundamental lemma, we formulate FRF estimation as a DRO problem, wherein robustness is desired with respect to the unknown probability distribution generating the noisy data. The proposed estimator computes FRF estimates by minimizing a convex upper bound on the DRO cost and allows for frequency-dependent calibration of the Wasserstein metric; the tractable upper bound consists of an empirical fit and a regularizer that penalizes the largest weighted sensitivity to spectral perturbations. Under suitable assumptions, we prove that the FRF estimation error is upper bounded by the optimal value of the DRO problem. The method is validated via a simple numerical study, which demonstrates improved performance under severely heavy-tailed noise.

eess.SY↗