Adaptive Optics: Control and Calibration

This thesis investigates control and calibration methodologies for astronomical Adaptive Optics (AO) systems, focusing on three primary contributions: the development of an open-source real-time control software, enhancements in next-generation AO systems, and improvements to control strategies through machine learning. The goal of this research is to increase the accessibility, performance, and efficiency of AO systems by introducing new control architectures, scalable real-time solutions, and advanced calibration methods. The first contribution is pyRTC, an open-source software designed for kHz-speed real-time control in high actuator count AO systems. By leveraging Python for its high-level programmability, pyRTC demonstrates that real-time control tasks, traditionally confined to low-level environments, can be effectively managed within a flexible, high-level programming framework. Key features include multi-processing to bypass Python’s GIL, just-in-time compilation, GPU acceleration, and shared memory-based data transfer pipelines. pyRTC also integrates with PyTorch, allowing the streamlined testing of AI-based controllers in real-time. The second contribution focuses on the MMTO Adaptive exoPlanet characterization System (MAPS), in particular, the commissioning of a NIR PyWFS camera and the ESCAPE calibration protocol. ESCAPE, a hybrid approach combining empirical and pseudo-synthetic techniques, addresses calibration challenges unique to convex adaptive secondary mirrors, where internal light sources are inaccessible. Laboratory results confirm the effectiveness of ESCAPE for interaction matrix generation, while initial on-sky tests reveal limitations related to pupil illumination and hardware imperfections, which continue to be investigated. The final contribution explores AI-enhanced controllers, implemented on the RAZOR test bench, using a convolutional LSTM-based predictive controller. Initial results indicate that the controller significantly improves performance under high temporal error conditions. pyRTC facilitates real-time integration of these AI-based control methods, demonstrating the feasibility of implementing machine learning techniques in AO control systems. This thesis aligns with the projected needs of next-generation observatories as outlined in the Astro 2020 decadal survey, which anticipates an increased demand for AO instrumentation and a growing AO workforce. The findings contribute to the development of scalable, accessible AO solutions, aiming to support future observatories and foster advancements in astronomical AO research.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.