| Abstract: |
| In this talk, I will discuss insensitization problems for linear controlled evolutions under small parameter variations. The goal is to design a control which both drives the nominal system to a prescribed target and cancels the first-order sensitivity of the final state with respect to the parameter. I will present an abstract framework, in which this question is recast as a controllability problem for a coupled state-sensitivity cascade system. In finite dimension, I will give a complete characterization in terms of a Kalman-type rank condition. I will also illustrate the approach on PDE examples, including the heat equation with uncertain diffusion and the wave equation with uncertain potential. |
|