Assess optimization robustness in dynamically varying and noisy environments

Assess the effects of temporal fluctuations and measurement noise on the robustness of in-situ time-domain adjoint optimization when the physical system evolves during the optimization process.

Background

The proposed in-situ time-domain adjoint optimization protocol obtains gradients from experimentally measured forward and adjoint responses rather than from a reconstructed numerical model. Consequently, temporal fluctuations in the physical system and experimental noise may affect the consistency of the measured gradients and the convergence of the optimization loop.

The paper identifies operation in dynamically varying or noisy environments as a current limitation and states that the impact of these effects on optimization robustness has not yet been systematically assessed. Resolving this issue is important for extending the method to physical systems whose properties change during optimization.