Machine Learning extreme AO control for Exoearth detection

Terrestrial Exoplanets around very nearby M-stars within about 20 lightyears can be observed with the ELT Planetary Camera and Spectrograph (PCS) instrument to look for signs of extraterrestrial life. For this, PCS will need a superb eXtreme Adaptive Optics (XAO) system. The XAO error budget is dominated by the temporal delay error and photon noise. By averaging multiple past measurements and by overcoming the temporal delay between the present wavefront and the correction, predictive control promises to suppress both errors and make the observation of Exoearths possible.


The proposed PhD project pushes the idea of predictive control for XAO using Machine Learning (ML) techniques. This project will support the PCS technology roadmap within the Technology development Programme for which ESO already has ongoing collaborations with external institutes (ETH Zurich / CH, LUT Lappeenranta / FI). The goal of these collaborations is the development of appropriate algorithms through numerical simulations. Promising algorithms should then be tested in the laboratory and may eventually be brought on-sky with SCExAO (Subaru Telescope, Mauna Kea) and/or VLT-SPHERE. The laboratory setup (GPU-based High-order adaptive OpticS Testbench – GHOST) is currently under development and will become available early 2021.


The PhD candidate will address the experimental part of the work (commissioning of GHOST, experiment design, measurements, data analysis, presentation of results), and may also become involved in the on-sky testing if time permits. Specifically, the candidate will

  • be exposed to all major technologies used in adaptive optics instruments such as the deformable mirror, the wavefront sensor camera and the RTC, and to the scientific requirements for Exoplanet observations by high-contrast imaging instrumentation,
  • define the parameter space to be explored and prepare the setup accordingly. This will include the collection and analysis of on-sky turbulence data with VLT-SPHERE or the AOF, which can be recorded in parallel to any astronomical observation,
  • carry out experimental measurements with GHOST and possibly on-sky instrumentation
  • analyse the data and develop advanced signal processing skills,
  • disseminate the results and develop the ability for scientific writing.