Multi Kernel Gaussian Process Guiding: Predictive Correction for Mounts with Several Gear Stages

Abstract

Gaussian process (GP) based predictive periodic error correction, introduced by Klenske et al., learns a telescope mount's periodic tracking error online and applies the correction before the error is observed, removing the reaction lag inherent in purely reactive autoguiders. The published model uses a single periodic kernel and therefore represents the error of one gear stage: any repeating shape at one period, harmonics included, but only one period. Mounts with more than one reduction stage…