-
AUV trajectory optimization with hydrodynamic forces for Icy Moon Exploration
Authors:
Lukas Rust,
Shubham Vyas,
Bilal Wehbe
Abstract:
To explore oceans on ice-covered moons in the solar system, energy-efficient Autonomous Underwater Vehicles (AUVs) with long ranges must cover enough distance to record and collect enough data. These usually underactuated vehicles are hard to control when performing tasks such as vertical docking or the inspection of vertical walls. This paper introduces a control strategy for DeepLeng to navigate…
▽ More
To explore oceans on ice-covered moons in the solar system, energy-efficient Autonomous Underwater Vehicles (AUVs) with long ranges must cover enough distance to record and collect enough data. These usually underactuated vehicles are hard to control when performing tasks such as vertical docking or the inspection of vertical walls. This paper introduces a control strategy for DeepLeng to navigate in the ice-covered ocean of Jupiter's moon Europa and presents simulation results preceding a discussion on what is further needed for robust control during the mission.
△ Less
Submitted 14 June, 2024;
originally announced June 2024.
-
NIRPS first light and early science: breaking the 1 m/s RV precision barrier at infrared wavelengths
Authors:
Étienne Artigau,
François Bouchy,
René Doyon,
Frédérique Baron,
Lison Malo,
François Wildi,
Franceso Pepe,
Neil J. Cook,
Simon Thibault,
Vladimir Reshetov,
Xavier Dumusque,
Christophe Lovis,
Danuta Sosnowska,
Bruno L. Canto Martins,
Jose Renan De Medeiros,
Xavier Delfosse,
Nuno Santos,
Rafael Rebolo,
Manuel Abreu,
Guillaume Allain,
Romain Allart,
Hugues Auger,
Susana Barros,
Luc Bazinet,
Nicolas Blind
, et al. (89 additional authors not shown)
Abstract:
The Near-InfraRed Planet Searcher or NIRPS is a precision radial velocity spectrograph developed through collaborative efforts among laboratories in Switzerland, Canada, Brazil, France, Portugal and Spain. NIRPS extends to the 0.98-1.8 $μ$m domain of the pioneering HARPS instrument at the La Silla 3.6-m telescope in Chile and it has achieved unparalleled precision, measuring stellar radial velocit…
▽ More
The Near-InfraRed Planet Searcher or NIRPS is a precision radial velocity spectrograph developed through collaborative efforts among laboratories in Switzerland, Canada, Brazil, France, Portugal and Spain. NIRPS extends to the 0.98-1.8 $μ$m domain of the pioneering HARPS instrument at the La Silla 3.6-m telescope in Chile and it has achieved unparalleled precision, measuring stellar radial velocities in the infrared with accuracy better than 1 m/s. NIRPS can be used either stand-alone or simultaneously with HARPS. Commissioned in late 2022 and early 2023, NIRPS embarked on a 5-year Guaranteed Time Observation (GTO) program in April 2023, spanning 720 observing nights. This program focuses on planetary systems around M dwarfs, encompassing both the immediate solar vicinity and transit follow-ups, alongside transit and emission spectroscopy observations. We highlight NIRPS's current performances and the insights gained during its deployment at the telescope. The lessons learned and successes achieved contribute to the ongoing advancement of precision radial velocity measurements and high spectral fidelity, further solidifying NIRPS' role in the forefront of the field of exoplanets.
△ Less
Submitted 13 June, 2024; v1 submitted 12 June, 2024;
originally announced June 2024.
-
NIRPS Front-End: Design, performance, and lessons learned
Authors:
N. Blind,
U. Conod,
A. de Meideros,
F. Wildi,
F. Bouchy,
S. Bovay,
D. Brousseau,
A. Cabral,
L. Genolet,
J. Kolb,
R. Schnell,
A. Segovia,
M. Sordet,
S. Thibault,
B. Wehbe,
G. Zins
Abstract:
NIRPS (Near Infra-Red Planet Searcher) is an AO-assisted and fiber-fed spectrograph for high precision radial velocity measurements in the YJH-bands. NIRPS also has the specificity to be an SCAO assisted instrument, enabling the use of few-mode fibers for the first time. This choice offers an excellent trade-off by allowing to design a compact cryogenic spectrograph, while maintaining a high coupl…
▽ More
NIRPS (Near Infra-Red Planet Searcher) is an AO-assisted and fiber-fed spectrograph for high precision radial velocity measurements in the YJH-bands. NIRPS also has the specificity to be an SCAO assisted instrument, enabling the use of few-mode fibers for the first time. This choice offers an excellent trade-off by allowing to design a compact cryogenic spectrograph, while maintaining a high coupling efficiency under bad seeing conditions and for faint stars. The main drawback resides in a much more important modal-noise, a problem that has to be tackled for allowing 1m/s precision radial velocity measurements. In this paper, we present the NIRPS Front-End: an overview of its design (opto-mechanics, control), its performance on-sky, as well as a few lessons learned along the way.
△ Less
Submitted 28 July, 2022;
originally announced July 2022.
-
Spatial Acoustic Projection for 3D Imaging Sonar Reconstruction
Authors:
Sascha Arnold,
Bilal Wehbe
Abstract:
In this work we present a novel method for reconstructing 3D surfaces using a multi-beam imaging sonar. We integrate the intensities measured by the sonar from different viewpoints for fixed cell positions in a 3D grid. For each cell we integrate a feature vector that holds the mean intensity for a discretized range of viewpoints. Based on the feature vectors and independent sparse range measureme…
▽ More
In this work we present a novel method for reconstructing 3D surfaces using a multi-beam imaging sonar. We integrate the intensities measured by the sonar from different viewpoints for fixed cell positions in a 3D grid. For each cell we integrate a feature vector that holds the mean intensity for a discretized range of viewpoints. Based on the feature vectors and independent sparse range measurements that act as ground truth information, we train convolutional neural networks that allow us to predict the signed distance and direction to the nearest surface for each cell. The predicted signed distances can be projected into a truncated signed distance field (TSDF) along the predicted directions. Utilizing the marching cubes algorithm, a polygon mesh can be rendered from the TSDF. Our method allows a dense 3D reconstruction from a limited set of viewpoints and was evaluated on three real-world datasets.
△ Less
Submitted 6 June, 2022;
originally announced June 2022.
-
Self-supervised Learning for Sonar Image Classification
Authors:
Alan Preciado-Grijalva,
Bilal Wehbe,
Miguel Bande Firvida,
Matias Valdenegro-Toro
Abstract:
Self-supervised learning has proved to be a powerful approach to learn image representations without the need of large labeled datasets. For underwater robotics, it is of great interest to design computer vision algorithms to improve perception capabilities such as sonar image classification. Due to the confidential nature of sonar imaging and the difficulty to interpret sonar images, it is challe…
▽ More
Self-supervised learning has proved to be a powerful approach to learn image representations without the need of large labeled datasets. For underwater robotics, it is of great interest to design computer vision algorithms to improve perception capabilities such as sonar image classification. Due to the confidential nature of sonar imaging and the difficulty to interpret sonar images, it is challenging to create public large labeled sonar datasets to train supervised learning algorithms. In this work, we investigate the potential of three self-supervised learning methods (RotNet, Denoising Autoencoders, and Jigsaw) to learn high-quality sonar image representation without the need of human labels. We present pre-training and transfer learning results on real-life sonar image datasets. Our results indicate that self-supervised pre-training yields classification performance comparable to supervised pre-training in a few-shot transfer learning setup across all three methods. Code and self-supervised pre-trained models are be available at https://github.com/agrija9/ssl-sonar-images
△ Less
Submitted 20 April, 2022;
originally announced April 2022.
-
Deep Reinforcement Learning for Continuous Docking Control of Autonomous Underwater Vehicles: A Benchmarking Study
Authors:
Mihir Patil,
Bilal Wehbe,
Matias Valdenegro-Toro
Abstract:
Docking control of an autonomous underwater vehicle (AUV) is a task that is integral to achieving persistent long term autonomy. This work explores the application of state-of-the-art model-free deep reinforcement learning (DRL) approaches to the task of AUV docking in the continuous domain. We provide a detailed formulation of the reward function, utilized to successfully dock the AUV onto a fixe…
▽ More
Docking control of an autonomous underwater vehicle (AUV) is a task that is integral to achieving persistent long term autonomy. This work explores the application of state-of-the-art model-free deep reinforcement learning (DRL) approaches to the task of AUV docking in the continuous domain. We provide a detailed formulation of the reward function, utilized to successfully dock the AUV onto a fixed docking platform. A major contribution that distinguishes our work from the previous approaches is the usage of a physics simulator to define and simulate the underwater environment as well as the DeepLeng AUV. We propose a new reward function formulation for the docking task, incorporating several components, that outperforms previous reward formulations. We evaluate proximal policy optimization (PPO), twin delayed deep deterministic policy gradients (TD3) and soft actor-critic (SAC) in combination with our reward function. Our evaluation yielded results that conclusively show the TD3 agent to be most efficient and consistent in terms of docking the AUV, over multiple evaluation runs it achieved a 100% success rate and episode return of 10667.1 +- 688.8. We also show how our reward function formulation improves over the state of the art.
△ Less
Submitted 5 August, 2021;
originally announced August 2021.
-
Pre-trained Models for Sonar Images
Authors:
Matias Valdenegro-Toro,
Alan Preciado-Grijalva,
Bilal Wehbe
Abstract:
Machine learning and neural networks are now ubiquitous in sonar perception, but it lags behind the computer vision field due to the lack of data and pre-trained models specifically for sonar images. In this paper we present the Marine Debris Turntable dataset and produce pre-trained neural networks trained on this dataset, meant to fill the gap of missing pre-trained models for sonar images. We t…
▽ More
Machine learning and neural networks are now ubiquitous in sonar perception, but it lags behind the computer vision field due to the lack of data and pre-trained models specifically for sonar images. In this paper we present the Marine Debris Turntable dataset and produce pre-trained neural networks trained on this dataset, meant to fill the gap of missing pre-trained models for sonar images. We train Resnet 20, MobileNets, DenseNet121, SqueezeNet, MiniXception, and an Autoencoder, over several input image sizes, from 32 x 32 to 96 x 96, on the Marine Debris turntable dataset. We evaluate these models using transfer learning for low-shot classification in the Marine Debris Watertank and another dataset captured using a Gemini 720i sonar. Our results show that in both datasets the pre-trained models produce good features that allow good classification accuracy with low samples (10-30 samples per class). The Gemini dataset validates that the features transfer to other kinds of sonar sensors. We expect that the community benefits from the public release of our pre-trained models and the turntable dataset.
△ Less
Submitted 2 August, 2021;
originally announced August 2021.
-
On-sky measurements of atmospheric dispersion: II. Atmospheric models characterization
Authors:
B. Wehbe.,
A. Cabral,
L. Sbordone,
G. Avila
Abstract:
Differential atmospheric dispersion is a wavelength-dependent effect introduced by Earth's atmosphere that affects astronomical observations performed using ground-based telescopes. It is important, when observing at a zenithal angle different from zero, to use an Atmospheric Dispersion Corrector (ADC) to compensate this atmospheric dispersion. The design of an ADC is based on atmospheric models t…
▽ More
Differential atmospheric dispersion is a wavelength-dependent effect introduced by Earth's atmosphere that affects astronomical observations performed using ground-based telescopes. It is important, when observing at a zenithal angle different from zero, to use an Atmospheric Dispersion Corrector (ADC) to compensate this atmospheric dispersion. The design of an ADC is based on atmospheric models that, to the best of our knowledge, were never tested against on-sky measurements. We present an extensive models analysis in the wavelength range of 315-665 nm. The method we used was previously described in the paper I of this series. It is based on the use of cross-dispersion spectrographs to determine the position of the centroid of the spatial profile at each wavelength of each spectral order. The accuracy of the method is 18 mas. At this level, we are able to compare and characterize the different atmospheric dispersion models of interest. For better future ADC designs, we recommend to avoid the Zemax model, and in particular in the blue range of the spectra, when expecting residuals at the level of few tens of milli-arcseconds.
△ Less
Submitted 3 March, 2021;
originally announced March 2021.
-
A First Step Towards Distribution Invariant Regression Metrics
Authors:
Mario Michael Krell,
Bilal Wehbe
Abstract:
Regression evaluation has been performed for decades. Some metrics have been identified to be robust against shifting and scaling of the data but considering the different distributions of data is much more difficult to address (imbalance problem) even though it largely impacts the comparability between evaluations on different datasets. In classification, it has been stated repeatedly that perfor…
▽ More
Regression evaluation has been performed for decades. Some metrics have been identified to be robust against shifting and scaling of the data but considering the different distributions of data is much more difficult to address (imbalance problem) even though it largely impacts the comparability between evaluations on different datasets. In classification, it has been stated repeatedly that performance metrics like the F-Measure and Accuracy are highly dependent on the class distribution and that comparisons between different datasets with different distributions are impossible. We show that the same problem exists in regression. The distribution of odometry parameters in robotic applications can for example largely vary between different recording sessions. Here, we need regression algorithms that either perform equally well for all function values, or that focus on certain boundary regions like high speed. This has to be reflected in the evaluation metric. We propose the modification of established regression metrics by weighting with the inverse distribution of function values $Y$ or the samples $X$ using an automatically tuned Gaussian kernel density estimator. We show on synthetic and robotic data in reproducible experiments that classical metrics behave wrongly, whereas our new metrics are less sensitive to changing distributions, especially when correcting by the marginal distribution in $X$. Our new evaluation concept enables the comparison of results between different datasets with different distributions. Furthermore, it can reveal overfitting of a regression algorithm to overrepresented target values. As an outcome, non-overfitting regression algorithms will be more likely chosen due to our corrected metrics.
△ Less
Submitted 10 September, 2020;
originally announced September 2020.
-
On-sky measurements of atmospheric dispersion: I. Method validation
Authors:
B. Wehbe,
A. Cabral,
G. Avila
Abstract:
Observations with ground-based telescopes are affected by differential atmospheric dispersion due to the wavelength-dependent index of refraction of the atmosphere. The usage of an Atmospheric Dispersion Corrector (ADC) is fundamental to compensate this effect. Atmospheric dispersion correction residuals above the level of ~ 100 milli-arcseconds (mas) will affect astronomical observations, in part…
▽ More
Observations with ground-based telescopes are affected by differential atmospheric dispersion due to the wavelength-dependent index of refraction of the atmosphere. The usage of an Atmospheric Dispersion Corrector (ADC) is fundamental to compensate this effect. Atmospheric dispersion correction residuals above the level of ~ 100 milli-arcseconds (mas) will affect astronomical observations, in particular radial velocity and flux losses. The design of an ADC is based on atmospheric models. To the best of our knowledge, those models have never been tested on-sky. In this paper, we present a new method to measure the atmospheric dispersion on-sky in the optical range. We require an accuracy better than 50 mas that is equal to the difference between atmospheric models. The method is based on the use of cross-dispersion spectrographs to determine the position of the centroid of the spatial profile at each wavelength of each spectral order. The method is validated using cross-dispersed spectroscopic data acquired with the slit spectrograph UVES. We measure an instrumental dispersion of 47 mas in the blue arm, 15 mas, and 23 mas in the two ranges of the red arm. We also measure a 4 % deviation in the pixel scale from the value cited in UVES manual. The accuracy of the method is ~ 17 mas in the range of 315-665 nm. At this level, we can compare and characterize different atmospheric dispersion models for better future ADC designs.
△ Less
Submitted 3 September, 2020;
originally announced September 2020.
-
The impact of atmospheric dispersion in the performance of high-resolution spectrographs
Authors:
Bachar Wehbe,
Alexandre Cabral,
Jorge Martins,
Pedro Figueira,
Nuno Santos,
Gerardo Ávila
Abstract:
Differential atmospheric dispersion is a wavelength-dependent effect introduced by the atmosphere. It is one of the instrumental errors that can affect the position of the target as perceived on the sky and its flux distribution. This effect will affect the results of astronomical observations if not corrected by an atmospheric dispersion corrector (ADC). In high-resolution spectrographs, in order…
▽ More
Differential atmospheric dispersion is a wavelength-dependent effect introduced by the atmosphere. It is one of the instrumental errors that can affect the position of the target as perceived on the sky and its flux distribution. This effect will affect the results of astronomical observations if not corrected by an atmospheric dispersion corrector (ADC). In high-resolution spectrographs, in order to reach a radial velocity (RV) precision of 10 cm/s, an ADC is expected to return residuals at only a few tens of milli-arcseconds (mas). In fact, current state-of-the-art spectrographs conservatively require this level of residuals, although no work has been done to quantify the impact of atmospheric dispersion. In this work we test the effect of atmospheric dispersion on astronomical observations in general, and in particular on RV precision degradation and flux losses. Our scientific objective was to quantify the amount of residuals needed to fulfill the requirements set on an ADC during the design phase. We found that up to a dispersion of 100 mas, the effect on the RV is negligible. However, on the flux losses, such a dispersion can create a loss of ~2% at 380 nm, a significant value when efficiency is critical. The requirements set on ADC residuals should take into consideration the atmospheric conditions where the ADC will function, and also all the aspects related with not only the RV precision requirements but also the guiding camera used, the tolerances on the flux loss, and the different melt data of the chosen glasses.
△ Less
Submitted 19 November, 2019;
originally announced November 2019.
-
Results from the Robocademy ITN: Autonomy, Disturbance Rejection and Perception for Advanced Marine Robotics
Authors:
Matias Valdenegro-Toro,
Mariela De Lucas Alvarez,
Mariia Dmitrieva,
Bilal Wehbe,
Georgios Salavasidis,
Shahab Heshmati-Alamdari,
Juan F. Fuentes-Pérez,
Veronika Yordanova,
Klemen Istenič,
Thomas Guerneve
Abstract:
Marine and Underwater resources are important part of the economy of many countries. This requires significant financial resources into their construction and maintentance. Robotics is expected to fill this void, by automating and/or removing humans from hostile environments in order to easily perform maintenance tasks. The Robocademy Marie Sklodowska-Curie Initial Training Network was funded by t…
▽ More
Marine and Underwater resources are important part of the economy of many countries. This requires significant financial resources into their construction and maintentance. Robotics is expected to fill this void, by automating and/or removing humans from hostile environments in order to easily perform maintenance tasks. The Robocademy Marie Sklodowska-Curie Initial Training Network was funded by the European Union's FP7 research program in order to train 13 Fellows into world-leading researchers in Marine and Underwater Robotics. The fellows developed guided research into three areas of key importance: Autonomy, Disturbance Rejection, and Perception. This paper presents a summary of the fellows' research in the three action lines. 71 scientific publications were the primary result of this project, with many other publications currently in the pipeline. Most of the fellows have found employment in Europe, which shows the high demand for this kind of experts. We believe the results from this project are already having an impact in the marine robotics industry, as key technologies are being adopted already.
△ Less
Submitted 29 October, 2019;
originally announced October 2019.
-
The development of an optical design tool for atmospheric dispersion correction
Authors:
Bachar Wehbe,
Alexandre Cabral,
Gerardo Avila
Abstract:
In ground based astronomical observations, atmospheric dispersion shifts the image of the object at different wavelengths due to the wavelength-dependent index of refraction of the atmosphere. Thus, using an Atmospheric Dispersion Corrector (ADC) is mandatory in order to avoid any wavelength dependent losses. Typical ADC configurations, for high resolution astronomical instruments, are two counter…
▽ More
In ground based astronomical observations, atmospheric dispersion shifts the image of the object at different wavelengths due to the wavelength-dependent index of refraction of the atmosphere. Thus, using an Atmospheric Dispersion Corrector (ADC) is mandatory in order to avoid any wavelength dependent losses. Typical ADC configurations, for high resolution astronomical instruments, are two counter-rotating prisms, a set of, at least, four prisms paired together. With the arrival of large telescopes with higher angular magnification, and spectrographs with higher resolution, the requirements on the dispersion correction are becoming more critical due to the impact on the produced science (e.g. radial velocity precision). We developed an ADC optical design tool in order to select the best set of glasses in terms of residuals, transmission, resulting image quality, Fresnel losses, taking into account the required spectral range and typical atmospheric conditions where the ADC will be working. A demonstration of the capabilities of the tool is presented with the analysis of the impact of different melt data, the effect of different glass Sellmeier coefficients between catalog and measured ones, that can create a difference in the residuals above few tens of milli-arcseconds (mas). The tool allows the investigation of critical steps on the ADC design phase and speeds up the glass selection process critical for the harder requirements of the future instruments/telescopes.
△ Less
Submitted 8 October, 2019;
originally announced October 2019.
-
Atmospheric dispersion correction: model requirements and impact on radial velocity measurements
Authors:
Bachar Wehbe,
Alexandre Cabral,
Pedro Figueira,
Gerardo Avila
Abstract:
Observations with ground-based telescopes are affected by differential atmospheric dispersion when seen at a zenith angle different from zero, a consequence of the wavelength-dependent index of refraction of the atmosphere. One of the pioneering technology in detecting exoplanets is the technique of radial velocity (RV), that can be affected by uncorrected atmospheric dispersion. The current highe…
▽ More
Observations with ground-based telescopes are affected by differential atmospheric dispersion when seen at a zenith angle different from zero, a consequence of the wavelength-dependent index of refraction of the atmosphere. One of the pioneering technology in detecting exoplanets is the technique of radial velocity (RV), that can be affected by uncorrected atmospheric dispersion. The current highest precision spectrographs are expected to deliver a precision of 10 cm/s (e.g., ESPRESSO). To minimize the atmospheric dispersion effect, an Atmospheric Dispersion Corrector (ADC) can be employed. ADC designs are based on sky dispersion models that nonetheless give different results; these can reach a few tens of milli-arcseconds (mas) in the sky (a difference up to 40 mas); a value close to the current requirements (20 mas in the case of ESPRESSO). In this paper we describe tests done with ESPRESSO and HARPS to understand the influence of atmospheric dispersion and its correction on RV precision. We also present a comparison of different sky models, using EFOSC2 data (between 600nm and 700nm), that will be used to improve on the design of ADCs.
△ Less
Submitted 8 October, 2019;
originally announced October 2019.
-
A Framework for On-line Learning of Underwater Vehicles Dynamic Models
Authors:
Bilal Wehbe,
Marc Hildebrandt,
Frank Kirchner
Abstract:
Learning the dynamics of robots from data can help achieve more accurate tracking controllers, or aid their navigation algorithms. However, when the actual dynamics of the robots change due to external conditions, on-line adaptation of their models is required to maintain high fidelity performance. In this work, a framework for on-line learning of robot dynamics is developed to adapt to such chang…
▽ More
Learning the dynamics of robots from data can help achieve more accurate tracking controllers, or aid their navigation algorithms. However, when the actual dynamics of the robots change due to external conditions, on-line adaptation of their models is required to maintain high fidelity performance. In this work, a framework for on-line learning of robot dynamics is developed to adapt to such changes. The proposed framework employs an incremental support vector regression method to learn the model sequentially from data streams. In combination with the incremental learning, strategies for including and forgetting data are developed to obtain better generalization over the whole state space. The framework is tested in simulation and real experimental scenarios demonstrating its adaptation capabilities to changes in the robot's dynamics.
△ Less
Submitted 13 March, 2019;
originally announced March 2019.
-
Learning of Multi-Context Models for Autonomous Underwater Vehicles
Authors:
Bilal Wehbe,
Octavio Arriaga,
Mario Michael Krell,
Frank Kirchner
Abstract:
Multi-context model learning is crucial for marine robotics where several factors can cause disturbances to the system's dynamics. This work addresses the problem of identifying multiple contexts of an AUV model. We build a simulation model of the robot from experimental data, and use it to fill in the missing data and generate different model contexts. We implement an architecture based on long-s…
▽ More
Multi-context model learning is crucial for marine robotics where several factors can cause disturbances to the system's dynamics. This work addresses the problem of identifying multiple contexts of an AUV model. We build a simulation model of the robot from experimental data, and use it to fill in the missing data and generate different model contexts. We implement an architecture based on long-short-term-memory (LSTM) networks to learn the different contexts directly from the data. We show that the LSTM network can achieve high classification accuracy compared to baseline methods, showing robustness against noise and scaling efficiently on large datasets.
△ Less
Submitted 17 September, 2018;
originally announced September 2018.
