Ground control to major time-lag: on-sky results of data-driven predictive wavefront control at Keck Observatory

Abstract

Directly imaging and characterizing exoplanets requires extreme adaptive optics (XAO), which achieves exquisite wavefront correction over a small (<5") field of view. Temporal errors, where the wavefront evolves faster than the lag between wavefront sensing and control, are often a leading term in the error budget for these XAO systems. Predictive control mitigates temporal errors by predicting where the wavefront will be by the time the system correction is applied. In particular, empirical orthogonal functions (EOF) learn linear correlations in a wavefront using previous states in the wavefront sensor history. We present on-sky results of a new implementation of EOF built directly into the Keck-II real time controller. On-sky engineering tests at Keck Observatory of the predictive controller show a 20% performance improvement over a classic integrator according to wavefront residuals from the Shack-Hartmann Wavefront Sensor (SHWFS). Parameter optimization studies show that there is a clear improvement based on varying predictive filter hyper-parameters, but that within a reasonable regime, varying filter parameters does not degrade performance to notably worse than an integrator. NIRC2 imaging through the Brackett Gamma=2190nm filter shows comparable performance between an integrator and predictor, both comparing Strehl Ratio (SR) and coronagraph-free contrast. We also explore power in principal components, and find a modest improvement (on the order of 3% less area under the curve of component strength) from the predictor over the integrator. This work not only improves current observing for the Keck community, but also acts as a pathfinder for predictive control methods with extremely large telescopes.