Search

Your search keyword '"Rico A"' showing total 1,765 results
Start Over Author "Rico A" Publisher ieee
1,765 results on '"Rico A"'

55. Optimizing Distributed Deep Learning in Heterogeneous Computing Platforms for Remote Sensing Data Classification

Author

Moreno-Alvarez, Sergio, Paoletti, Mercedes E., Rico, Juan A., Cavallaro, Gabriele, and Haut, Juan M.

Abstract

Applications from Remote Sensing (RS) unveiled unique challenges to Deep Learning (DL) due to the high volume and complexity of their data. On the one hand, deep neural network architectures have the capability to automatically ex-tract informative features from RS data. On the other hand, these models have massive amounts of tunable parameters, requiring high computational capabilities. Distributed DL with data parallelism on High-Performance Computing (HPC) systems have proved necessary in dealing with the demands of DL models. Nevertheless, a single HPC system can be al-ready highly heterogeneous and include different computing resources with uneven processing power. In this context, a standard data parallelism strategy does not partition the data efficiently according to the available computing resources. This paper proposes an alternative approach to compute the gradient, which guarantees that the contribution to the gradient calculation is proportional to the processing speed of each DL model's replica. The experimental results are obtained in a heterogeneous HPC system with RS data and demonstrate that the proposed approach provides a significant training speed up and gain in the global accuracy compared to one of the state-of-the-art distributed DL framework.

Published
2022

74. Feedforward for Robust Reference Tracking in Multi-Input Feedback Control

Author

Rico-Azagra, Javier, Gil-Martinez, Montserrat, 0000-0002-5500-6093, and 0000-0002-6547-5301

Subjects
Frequency response, General Computer Science, quantitative feedback theory, Computer science, Bandwidth (signal processing), Automatic frequency control, General Engineering, Feed forward, control allocation, model-matching, TK1-9971, Reduction (complexity), Noise, Quantitative feedback theory, Mid-ranging control, Control theory, General Materials Science, Electrical engineering. Electronics. Nuclear engineering, Robust control, valve position control, robust control
Abstract

This paper examines methods to incorporate feedforward loops of known external inputs (output reference) into a multi-input feedback control structure to achieve certain robust performance of its output. Undoubtedly, feedforward can reduce the need for feedback and therefore the amplification of sensor noise at actuators, as occurs in single-input control. Beyond that, since there are several available inputs, a convenient distribution of feedforward and feedback can minimise the control action at each input and offer benefits at all frequencies. The procedure is as follows: because there are rough plant models of the behaviour from each input to the output, it is possible to approximate the individual control demand that will satisfy the performance. Based on this, individual feedforward filters allocate the control bandwidth among the inputs in order to build an equivalent plant that has an equal or greater magnitude than any individual plant at each frequency. Next, the uncertainty of this equivalent plant is addressed by feedback that reduces the closed loop deviation of magnitude frequency responses. The reduction is sufficient to enable a master feedforward to place them, at a second step, around the desired tracking performance model without violating any deviation tolerances. Individual feedback controllers distribute the total feedback among the inputs in order to have the least possible feedback at each frequency. A first example illustrates the method and the relevance of a feedforward orientation to reduce the individual control action, instead of the individual feedback action. Another example proves the superiority of adding feedforward loops to feedback-only schemes and highlights the benefits of robust design methods such as Quantitative Feedback Theory (QFT). This paper also provides the algorithms to employ in response to new robust control specifications in the framework of QFT.

Published
2021

87. Obstacle Coordinates Transformation from TVS Body-Frame to AGV Navigation-Frame

Author

Garcia-Gutierrez, Xavier de J., primary, Alaniz-Plata, Ruben, additional, Trujillo-Hernandez, Gabriel, additional, Alba-Corpus, Ivan Y., additional, Flores-Fuentes, Wendy, additional, Rodriguez-Quinonez, Julio C., additional, Hernandez-Balbuena, Daniel, additional, Sergiyenko, Oleg, additional, Gonzalez-Navarro, Felix F., additional, Mercorelli, Paolo, additional, Miranda-Vega, Jesus E., additional, and Murrieta-Rico, Fabian N., additional

Published
2022
Full Text
View/download PDF

94. Direct Measurement of Neutral Gas during Hypervelocity Planetary Flybys

Author

Fausch, Rico G., Wurz, Peter, Cotting, Björn, Rohner, Urs, and Tulej, Marek

Subjects
530 Physics, 520 Astronomy, 620 Engineering
Abstract

Planetary flybys at hypervelocity provide the opportunity for sensitive, in situ chemical composition analysis performed with mass spectrometers to identify both simple and complex species in search of the origin and evolution of life in the Solar System. Current mass spectrometric instrumentation handles the fast flow of neutral gas (and ions) streaming into the instrument with an open or closed ion source system. In an open source system, the particles enter the ion source directly without contact with any element, are ionized there, and then are deflected into the mass analyzer section by electrostatic means. The advantage of an open source is that there is no particle alteration by surface interactions, but the electrostatic deflection limits the upper velocity of incoming particles to about 5 km/s. In a closed source system, the particles enter an antechamber through a small hole first. Many collisions with the chamber wall thermalize them before they enter the actual ion source as thermalized gas. The advantages of a closed source system include the large velocity range and the increased sensitivity resulting from the ram pressure enhancement. Its disadvantage is that the particles may fragment, or chemically alter as they hit the wall leading to complicated fractionation patterns in the recorded mass spectrum. Thus, the larger the molecules are the more difficult, or even impossible, inferring the original molecule becomes. Here we present a novel ion optical system for a mass spectrometer that directly measures the atmospheric species at orbital velocities with an open source system, but without ion deflection into the mass analyzer section to overcome the disadvantages of both techniques and unite their advantages. Furthermore, the mass analyzer features a selectivity for the velocity of the analyzed gas allowing for the first time for sensitive mass spectrometric analysis of the atmospheric gas, while reliably knowing the influence of the background gas resulting from spacecraft outgassing. As a first realization, we built an ion optical system for 1U of a CubeSat (CubeSatTOF). Our results demonstrate that the sample acquisition system preserves the chemical nature of molecules even when relative encounter velocities of up to 20 km/s are required. Its mass range is about m/z 1 to 300, thanks to its mass resolution exceeding 300 (FWHM), designed for an application of analyzing Earth's exosphere. In addition, we designed a larger system (OpenTOF) with a mass range of about m/z 1 to 1,000 and a mass resolution of up to 1,000 (FWHM) for flybys of Enceladus, Io, Europa, Ganymede, Venus, and others, capable of analysis of complex (bio) molecules.

Published
2022

100. CHESS: Measuring the Dynamics of Composition and Density of Earth's Upper Atmosphere with CubeSats

Author

Fausch, Rico G., primary, Moeller, Gregor, additional, Rothacher, Markus, additional, Martinod, Nicolas, additional, Trebaol, Tristan, additional, Villegas, Alfonso, additional, Kneib, Jean-Paul, additional, Corthay, Francois, additional, Joss, Marcel, additional, Tieche, Francois, additional, Tulej, Marek, additional, and Wurz, Peter, additional

Published
2022
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results