Using artificial neural networks for open-loop tomography

Osborn, James; Juez, Francisco Javier de Cos; Guzmán, Dani; Butterley, T.; Myers, R; Guesalaga, Andrés; Laine, Jesus

doi:10.1364/oe.20.002420

Cited by 50 publications

(36 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The SLODAR can be adapted to use LGSs as previously reported (Fusco & Costille ; Gilles & Ellerbroek ; Osborn et al ). In such cases, the turbulence profiling with LGSs is performed to non‐equally spaced bin altitudes due to the cone effect.…”

Section: The Slodar Methodsmentioning

confidence: 99%

Atmospheric turbulence profiling using multiple laser star wavefront sensors

Cortes

Neichel

Guesalaga

et al. 2012

Monthly Notices of the Royal Astronomical Society

Self Cite

View full text Add to dashboard Cite

This paper describes the data pre-processing and reduction methods together with SLOpe Detection And Ranging (SLODAR) analysis and wind profiling techniques for the Gemini South Multi-Conjugate Adaptive Optics System (GeMS).The wavefront gradient measurements of the five GeMS Shack-Hartmann sensors, each pointing to a laser guide star, are combined with the deformable mirror (DM) commands sent to three DMs optically conjugated at 0, 4.5 and 9 km in order to reconstruct pseudo-open loop slopes.These pseudo-open loop slopes are then used to reconstruct atmospheric turbulence profiles, based on the SLODAR and wind-profiling methods. We introduce the SLODAR method, and how it has been adapted to work in a closed-loop, multi-laser guide star system. We show that our method allows characterizing the turbulence of up to 16 layers for altitudes spanning from 0 to 19 km. The data pre-processing and reduction methods are described, and results obtained from observations made in 2011 are presented. The wind profiling analysis is shown to be a powerful technique not only for characterizing the turbulence intensity, wind direction and speed, but also as it can provide a verification tool for SLODAR results. Finally, problems such as the fratricide effect in multiple laser systems due to Rayleigh scattering, centroid gain variations, and limitations of the method are also addressed.

show abstract

Section: The Slodar Methodsmentioning

confidence: 99%

Atmospheric turbulence profiling using multiple laser star wavefront sensors

Cortes

Neichel

Guesalaga

et al. 2012

Monthly Notices of the Royal Astronomical Society

Self Cite

View full text Add to dashboard Cite

show abstract

“…N and wind speed profiles are retrieved using a method described in Neichel et al, 2014[ 2 ] using the SLODAR technique conceived for binaries (Wilson, 2002[ 23 ]) extended to multiple laser guide stars WFAO systems running with Shack-Hartmann wave-front sensors [ 1,24,25 ]. Turbulence strength is sampled with a vertical resolution ∆h = d/θ where d is the sub-aperture projected on the pupil and θ the angular separation of different couple of stars selected among the guide stars of the asterisms taken in consideration.…”

Section: Wind Speed From Gemsmentioning

confidence: 99%

Towards an automatic system for monitoring of C_N²and wind speed profiles with GeMS

et al. 2016

View full text Add to dashboard Cite

Wide Field Adaptive Optics (WFAO) systems represent the more sophisticated AO systems available today at large telescopes. One critical aspect for these WFAO systems in order to deliver an optimised performance is the knowledge of the vertical spatiotemporal distribution of the C 2 N and the wind speed. Previous studies (Cortes et al., 2012[ 1 ]) already proved the ability of GeMS (the Gemini Multi-Conjugated AO system) in retrieving C 2 N and wind vertical stratification using the telemetry data. To assess the reliability of the GeMS wind speed estimates a preliminary study [ 2 ]) compared wind speed retrieved from GeMS with that obtained with the atmospherical model Meso-Nh on a small sample of nights providing promising results. The latter technique is very reliable for the wind speed vertical stratification. The model outputs gave, indeed, an excellent agreement with a large sample of radiosoundings (∼ 50) both in statistical terms and on individual flights (Masciadri et al., 2013[ 3 ]). Such a tool can therefore be used as a valuable reference in this exercise of cross calibrating GeMS on-sky wind estimates with model predictions. The main results of analysis showed that, on a great number of cases, GeMS could reconstruct very good wind speed estimates. At the same time it has been put in evidence, on a number of cases, not negligible discrepancies from the atmospherical model. However we observed that these discrepancies strongly decreased or even disappear if GeMS data reduction is done with the a priori knowledge of the wind speed stratification provided by the model Meso-Nh. Basically the a priori knowledge helped the data reduction of GeMS acquisitions. In this contribution we achieved a two-fold results:(1) we extended analysis on a much richer statistical sample (∼ 43 nights), we confirmed the preliminary results and we found an even better correlation between GeMS observations and the atmospherical model with basically no cases of not-negligible uncertainties; (2) we evaluate the possibility to use, as an input for GeMS, the MesoNh estimates of the wind speed stratification in an operational configuration. Under this configuration these estimates can be provided many hours in advanced with respect to the observations and with a very high temporal frequency (order of 2 minutes or less). Such a system would have a set of advantages: (a) to implement inside GeMS a total temporal and spatial coverage of the wind speed over ∼ 20 km and all along the night not only in real-time but in advance of a few hours, (b) to improve the detection of the C 2 N vertical stratification from GeMS because a good wind speed estimation would improve the quality of the cross-correlation peaks detection, (c) the possibility to bypass the complex (and not necessarily reliable) procedures necessary to automatise the wind speed estimate of GeMS due to the relatively low vertical resolution of the system. Such a study can obviously be considered as a demonstrator for multiple operational AO and WFAO systems (AOF, L...

show abstract

“…Some of the most popular are based on a matrix vector multiplication, with the control matrix being defined by either least squares (LS) [7,8] or minimum variance techniques [9]. However, during recent years most complex solutions have been developed, like Learn and Apply (L&A) [10], or the intelligent system known as Complex Atmospheric Reconstructor based on Machine Learning (CARMEN) [11,12]. Due to the increasing complexity and amount of data used by these algorithms [13], some of previous algorithms have been implemented in Graphics Processing Units (GPUs) [14,15], speeding up substantially their execution and development [16,17].…”

Section: Introductionmentioning

confidence: 99%

Multi‐GPU Development of a Neural Networks Based Reconstructor for Adaptive Optics

et al. 2018

Self Cite

View full text Add to dashboard Cite

Aberrations introduced by the atmospheric turbulence in large telescopes are compensated using adaptive optics systems, where the use of deformable mirrors and multiple sensors relies on complex control systems. Recently, the development of larger scales of telescopes as the E-ELT or TMT has created a computational challenge due to the increasing complexity of the new adaptive optics systems. The Complex Atmospheric Reconstructor based on Machine Learning (CARMEN) is an algorithm based on artificial neural networks, designed to compensate the atmospheric turbulence. During recent years, the use of GPUs has been proved to be a great solution to speed up the learning process of neural networks, and different frameworks have been created to ease their development. The implementation of CARMEN in different Multi-GPU frameworks is presented in this paper, along with its development in a language originally developed for GPU, like CUDA. This implementation offers the best response for all the presented cases, although its advantage of using more than one GPU occurs only in large networks.

show abstract

Using artificial neural networks for open-loop tomography

Cited by 50 publications

References 25 publications

Atmospheric turbulence profiling using multiple laser star wavefront sensors

Atmospheric turbulence profiling using multiple laser star wavefront sensors

Towards an automatic system for monitoring of C_N²and wind speed profiles with GeMS

Multi‐GPU Development of a Neural Networks Based Reconstructor for Adaptive Optics

Contact Info

Product

Resources

About

Using artificial neural networks for open-loop tomography

Cited by 50 publications

References 25 publications

Atmospheric turbulence profiling using multiple laser star wavefront sensors

Atmospheric turbulence profiling using multiple laser star wavefront sensors

Towards an automatic system for monitoring of CN2and wind speed profiles with GeMS

Multi‐GPU Development of a Neural Networks Based Reconstructor for Adaptive Optics

Contact Info

Product

Resources

About

Towards an automatic system for monitoring of C_N²and wind speed profiles with GeMS