Electron Beam Improvements in Preparation for AWAKE Run 2: Human and Machine Learning

The AWAKE experiment accelerates externally injected electrons in plasma wakefields driven by a proton bunch from the CERN SPS.\footnote{AWAKE Collaboration, Nature 561, 363 (2018)} Improvements to the 18~MeV electron beam\footnote{C. Bracco et al., Proceedings of IPAC, 2019}, aimed at achieving parameters required for seeding the self-modulation of a long proton bunch in plasma, are in progress. First, we use standard techniques to improve beam parameters, including control and prediction of position and transverse properties at the plasma entrance\footnote{F. Pe\~{n}a et al., Proceedings of EAAC, 2019}, and to refine models used in predicting wakefields generated by different bunches. Second, we explore model-independent machine learning techniques to automatize and speed up the initial setup process, and to continuously react to external changes.\footnote{F. Velotti et al., paper in preparation}$^{,}$ \footnote{V. Kain et al., paper in preparation}$^{,}$ \footnote{A. Scheinker et al., AIP Advances 10, 055320 (2020)} We will present an overview of the 18~MeV electron beamline as well as our latest beam optimization and automation results.