|
Deformable mirrors are a keystone of adaptive optics, and flattening them is required to ensure a proper baseline for optical observations. This talk will go over developing a possible new algorithm to flatten the DM in a more efficient manner. In particular, the algorithm will attempt to use a database of control matrices and learned pre-determined phase screens to determine the best matrix for flattening multiple actuators simultaneously, as opposed to current methods of flattening each actuator one-by-one. I will discuss using a Python simulation of a 4f optical system with a DM and spatial light modulator (SLM), combined with reinforcement learning to form a DM-flattening algorithm, and testing that in a laboratory set-up. |
|