Integration of Data-driven Predictive Control into the new Keck-II Real Time Controller

From the ground, the direct imaging of exoplanets requires extreme adaptive optics (XAO), where the adaptive optics correction is optimized over a small
field-of-view. Keck Observatory has a recent real-time controller (RTC) upgrade and i s on the cusp of HAKA (a high order deformable mirror) integration;
this leaves bandwidth error - an error term driven by system lag time, as one of the few controllable error terms that stands between Keck AO and high
quality XAO performance. A natural solution to address bandwidth error is predictive wavefront control, where a predictive algorithm estimates the wavefront
shape at some future time when the correction will be applied. Predictive control improves low spatial frequency errors that impact science at small inner
working angles and increases sky coverage with the ability to run slower on fainter targets while only requiring a software change. A previous implementation
of a data-driven predictive control algorithm (specifically empirical orthogonal functions - EOF) at Keck showed an improvement in contrast of up to 3 for
separations of 3-7 X/D. However, this previous implementation was tied to an RTC associated with the IR pyramid wavefront sensor on the K-II AO bench,
which has since been decommissioned. With a new implementation of EOF on the Keck-II daytime bench, we present the first results of predictive control
at Keck with the new RTC, including performance improvements of up to 1.7 on the Keck-II AO bench and performance improvements from a preliminary
demonstration of closed-loop on sky. Facilitizing predictive control for Keck not only improves the AO quality for thousands of Keck users, but serves as
technology maturation for future extremely large telescopes.