The Compute and Control for Adaptive Optics (CACAO) is a free and open-source real-time library for adaptive optics (AO), initially developed for the operation of the 1200+ mode AO loop of Subaru/SCExAO. The scope has expanded since then, through refactorings, the addition of numerous features (predictive control, machine learning), and a substantial improvement of our understanding and configuration of the underlying real-time Linux distribution. We now witness the adoption of the package at multiple facilities, using a variety of cameras and WFSs: non-linear curvature, Shack-Hartmann, Photonic lanterns, and of course the pyWFS. At Subaru, CACAO is the core of the AO3K RTC, which supports legacy NGS and LGS mode, as well as the new high-order wavefront sensors coupled to an ALPAO 3224 deformable mirror. We present developments in algorithms -- bindings for machine learning algorithms, real-time configuration tools -- and user interface tools added in the past few years. We show performance benchmarks on the SCExAO and AO3K systems. We present our future plans to affirm CACAO as the go-to free, open-source RTC toolkit for real-time pipelines in the academic world.
The COSMIC platform endeavors to establish a potent, versatile, and accessible AO RTC platform for the AO community. This joint project allies the Observatoire de Paris with the Australian National University (ANU), and relies on the CACAO software, a creation of the Subaru Telescope. Although the primary intended use was for the ELT AO instrument, the platform now accommodates several systems on different scales. The effectiveness of COSMIC in current and future AO systems has already been proven. The current progress on the COSMIC platform extends beyond, with technical solutions offered for various applications other than AO such as embedded systems, radio astronomy, and radar. This interest exhibited by other communities has initiated new collaborations and additions to the platform, which might appeal to the AO community. This paper presents the current state of the COSMIC platform, including ongoing developments and future perspectives. One recent development is heterogeneous computation which allows for a hardware-agnostic real-time pipeline, thereby enabling the deployment of an AO pipeline on both CPU and GPU, irrespective of the vendor (NVIDIA or AMD). Another contribution involves the porting of COSMIC onto an NVIDIA Jetson Xavier, effectively paving the way for embedded system applications. As artificial intelligence is rapidly gaining importance in the AO community, efforts have been made to incorporate machine learning inference into the real-time critical path. Additionally, a collaboration with the radio astronomy community has led to a significant contribution towards efficient real-time data ingestion at a high volume. This paper aims to illustrate these achievements and demonstrate their potential applications within the AO community.
We present results on integrating Machine Learning (ML) methods for adaptive optics control with a real-time control library: COmmon Scalable and Modular Infrastructure for real-time Control (COSMIC). We test the integration on simulations for the instrument SAXO+. Our proposed solution’s pipeline is formed by a two-model ML system. The first model consists of a very Deep Neural Network (DNN) that maps Wavefront Sensor (WFS) images to phase and is trained offline. The second model consists of predictive control with a more compact DNN. The predictive control stage is trained online, providing an adaptive solution to changing atmospheric conditions but adding extra complexity to the pipeline. On top of implementing the solution with COSMIC, we add a set of modifications to provide faster inference and online training. Specifically, we test NVIDIA’s TensorRT to accelerate the DNNs inference, reduced precision, and just-in-time compilation for PyTorch. We show real-time capabilities by using COSMIC and improved speeds both in inference and training by using the recommendations mentioned above.
We present a model-free reinforcement learning (RL) predictive model with a supervised learning non-linear reconstructor for adaptive optics (AO) control with a pyramid wavefront sensor (P-WFS). First, we analyse the additional problems of training an RL control method with a P-WFS compared to the Shack-Hartmann WFS. From those observations, we propose our solution: a combination of model-free RL for prediction with a non-linear reconstructor based on neural networks with a U-net architecture. We test the proposed method in simulation of closed-loop AO for an 8m telescope equipped with a 32x32 P-WFS and observe that both the predictive and non-linear reconstruction add additional benefits over an optimised integrator.
A classical closed-loop adaptive optics system with a Shack-Hartmann wavefront sensor (WFS) relies on a center of gravity approach to process the WFS information and an integrator with gain to produce the commands to a Deformable Mirror (DM) to compensate wavefront perturbations. In this kind of systems, noise in the WFS images can propagate to errors in centroids computation, and thus, lead the AO system to perform poorly in closed-loop operations. In this work, we present a deep supervised learning method to denoise the WFS images based on convolutional denoising autoencoders. Our method is able to denoise the images up to a high noise level and improve the integrator performance almost to the level of a noise-free situation.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.