Hinfty-Optimal Observer Design for Linear Systems with Delays in States, Outputs and Disturbances

This paper considers the Hinfty-optimal estimation problem for linear systems with multiple delays in states, output, and disturbances. First, we formulate the Hinfty-optimal estimation problem in the Delay-Differential Equation (DDE) framework. Next, we construct an equivalent Partial Integral Equation (PIE) representation of the optimal estimator design framework. We then show that in the PIE framework, the Hinfty-optimal estimator synthesis problem can be posed as a Linear PI Inequality (LPI). LPIs are a generalization of LMIs to the algebra of Partial Integral (PI) operators and can be solved using the PIETOOLS toolbox. Finally, we convert the PIE representation of the optimal estimator back into an ODE-PDE representation - a form similar to a DDE, but with corrections to estimates of the infinite-dimensional state (the time-history). Numerical examples show that the synthesis condition we propose produces an estimator with provable Hinfty-gain bound which is accurate to 4 decimal places when compared with results obtained using Pade-based discretization.