Pasar al contenido principal

INTELIGENCIA ARTIFICIAL EN EL BORDE: SOLUCIONES EMBEBIDAS DE BAJO CONSUMO DE POTENCIA

PID2021-128009OB-C32

•
Nombre agencia financiadora Agencia Estatal de Investigación
Acrónimo agencia financiadora AEI
Programa Programa Estatal para Impulsar la Investigación Científico-Técnica y su Transferencia
Subprograma Subprograma Estatal de Generación de Conocimiento
Convocatoria Proyectos de I+D+i (Generación de Conocimiento y Retos Investigación)
Año convocatoria 2021
Plan Estatal 2021-2023 Plan Estatal de Investigación Científica y Técnica y de Innovación
Centro beneficiario UNIVERSIDADE DE SANTIAGO DE COMPOSTELA

Resultados relacionados

Resultados totales (Incluyendo duplicados): 10
Encontrada(s) 1 página(s)

Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Capítulo o parte de un libro (BookPart)

Live Demonstration: 5-bit signed SRAM-based DNN CIM for Image Recognition

  • Pereira Rial, Óscar
  • García Lesta, Daniel
  • Vaquero Otal, Lorenzo
  • López Martínez, Paula
  • Brea Sánchez, Víctor Manuel
  • Cabello Ferrer, Diego
This live demonstration shows a mixed-signal Computer In Memory (CIM) macro deep neural network (DNN) integrated circuit in 180 nm CMOS technology for image recognition. Images are coded as pulse width modulation (PWM) signals. DNN weights are stored as voltages in 6T-SRAM memories which drive current sources inside every multiplier. Multipliers are arranged within processing elements laid down in a 2D mesh suitable for image processing. The power consumption per multiplier of the CIM macro is of 0.22 µW, below state-of-the-art competitors following the same multiply and accumulate (MAC) principle.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Capítulo o parte de un libro (BookPart)

Single-Shot Transient Imaging Via Compressed Time-Of-Flight Imaging And Dictionary Learning

  • Fayyaz Shahandashti, Peyman
  • López Martínez, Paula
  • Brea Sánchez, Víctor Manuel
  • García-Lesta, Manuel
  • Heredia Conde, Miguel
In this work, we propose a single-shot transient imaging framework that combines sparsity-based signal modeling with compressive frequency-domain sampling. The temporal response function of a scene encodes rich information about depth and light transport, and its accurate reconstruction is critical for various imaging applications. To enable transient recovery from a limited number of measurements, we exploit learned sparse representations in an optimized dictionary basis. We compare multiple sampling strategies in the frequency domain and show that both transient profiles and depth maps can be reconstructed under highly compressed acquisition. Notably, we achieve full transient reconstruction using only 16 modulation frequencies, based on real correlation functions, enabling practical singleshot acquisition through spatial frequency multiplexing
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Contribución de congreso (ConferenceOutput)

Early vision on the focal-plane with high dynamic range pixels

  • Jaklin, Marko
  • García Lesta, Daniel
  • López Martínez, Paula
  • Brea Sánchez, Víctor Manuel
This paper introduces a high dynamic range pixel for early vision processing. Early vision is the first stage to subsequently extract semantic information for image processing or video analytics. This paper proposes to bring said processing to the focal plane, next to a high dynamic range image sensor working on the principle of lateral overflow capacitor. This brings the benefits of processing scenes with a wide dynamic range in a power efficient manner. Circuit simulations for edge detection, as an example of early vision processing conveyed in this paper, show that our proposal meets the accuracy typically found in applications like machine vision. Simulations are in XFAB’s XS018 technology
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Capítulo o parte de un libro (BookPart)

Maximum Output Power Point Tracking for Low Power Photovoltaic Energy Harvesting Systems

  • Vicente García, Laura
  • Pereira Rial, Óscar
  • López Martínez, Paula
A novel approach for finding the maximum output power point of a photovoltaic energy harvesting system is proposed in this paper. It is based on maximizing the power delivered to the load taking into account the performance of the whole system in contrast to conventional approaches focused on tracking the maximum power point of the photovoltaic transducer alone. The output power tracking is performed by measuring the output voltage, which significantly simplifies the hardware complexity and consequently reduces the power consumption. A circuit implementation of the entire system is presented and validated using electrical simulations. The designed circuit is self powered and is able to successfully supply a load with voltage levels higher than 1.2V. The system achieves a peak efficiency of 80.51% when the input power is equal to 28.03 μW.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Capítulo o parte de un libro (BookPart)

Piezoelectric Energy Harvesting Using Fully Integrated SECE Method: An Analytical Study of the Residual Voltage

  • Vicente García, Laura
  • Pereira Rial, Óscar
  • López Martínez, Paula
This article presents an analytical study of the residual voltage left in the piezoelectric capacitor after each cycle in inductor-less Synchronous Electric Charge Extraction (SECE) methods. A closed-form expression for this voltage is derived, and its impact on the energy harvesting efficiency is examined. The resulting equations are validated using numerical methods and spectre simulations. The analysis highlights key designs constraints, such as the DC-DC converter’s resistance, that limit energy extraction in fully integrated SECE-based harvesters.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Capítulo o parte de un libro (BookPart)

Recent Advances in Computational Time-of-Flight Imaging

  • Heredia Conde, Miguel
  • López Paredes, Álvaro
  • Ahmed, Faisal
  • Fayyaz Shahandashti, Peyman
  • López Martínez, Paula
Las cámaras de tiempo de vuelo (ToF) son dispositivos de imagen 3D que capturan la geometría de una escena, aprovechando que el tiempo de viaje de los fotones está directamente relacionado con la distancia recorrida. Este tipo de sensor es muy prometedor en áreas de aplicación emergentes, pero varias deficiencias dificultan su uso. En concreto, nos centramos en el consumo de energía relativamente alto en comparación con las cámaras convencionales, el rango limitado en el que se puede estimar la profundidad con precisión y las distorsiones de medición producidas por el contenido armónico de las formas de onda de modulación/demodulación y la interferencia multitrayecto. En este trabajo, presentamos avances recientes en la computación de imágenes ToF con el objetivo de superar estas limitaciones, permitiendo el funcionamiento pasivo en interiores, el funcionamiento a largo plazo en exteriores y la realización de operaciones de onda continua multifrecuencia de un solo disparo con mínima distorsión armónica. El funcionamiento pasivo se logra aprovechando fuentes de luz modulada, como los módulos LiFi o de Comunicaciones por Luz Visible (VLC). De forma independiente, se pueden obtener ganancias en el rango operativo mediante la conformación de pulsos ultracortos combinada con demodulación codificada de baja densidad. Se logra un muestreo de Fourier preciso con mínima distorsión armónica mediante la inducción de efectos de resonancia personalizados en los píxeles del tiempo de vuelo. Los resultados iniciales de la evaluación demuestran el potencial de estos enfoques computacionales de imágenes 3D para superar las limitaciones mencionadas.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Capítulo o parte de un libro (BookPart)

5-Bit Signed SRAM-Based In-Memory Computing Cell

  • Karimpour, Faranak
  • Pardo, F.
  • García Lesta, Daniel
Hardware accelerators are critical in providing real-time processing for edge computing applications, particu-larly in the context of convolutional neural networks. A crucial challenge in this context is achieving low power consumption while maintaining an appropriate performance in terms of accuracy. This work delves into a thorough analysis of prospective architectures for the core cell of the multiply -and -accumulate function, monitoring each structure's crucial benefits and drawbacks. It includes electrical simulations comparing their performance in a 180 nm process node for 1.8 V and 3.3 V. Moreover, a process corner simulation is proposed to identify on-chip process variations in the voltage error of the proposed design under different input voltages. Notably, the minimum corner errors observed at +15 and -7 sign bits are 0.45 % and 0.63 %, respectively. The significant outcome highlights that the single-switch implementation achieves optimal performance, displaying the lowest error value of 0.14%, specifically at the + 15 sign bit and operating at 1.8 V.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Artículo científico (JournalArticle)

Low-Voltage CMOS Capacitor-Less LDOs: Bulk-Driven Versus Gate-Driven Comparative Study

  • Pereira Rial, Óscar
  • Carrillo, Juan M.
  • López Martínez, Paula
This paper explores the feasibility of a capacitor-less (CL) low-dropout (LDO) regulator to operate efficiently in a low-voltage environment. The CL-LDO scheme selected is based on a unity-gain feedback configuration around the error amplifier (EA), so that the inclusion of high-value on-chip resistors is avoided and different key parameters, such as the power supply rejection or the noise, are optimized. A comparative analysis has been carried out over the same LDO structure including a bulk-driven and a gate-driven EA, respectively. The pass branch of the voltage regulator is provided with pseudo-class-AB operation, in order to lead to a very small quiescent current in the standby operation mode, whereas a very large current can be delivered to the load when required. Both regulators were designed and fabricated in 180 nm CMOS technology to operate with a maximum supply voltage of 1.8 V. The extensive experimental characterization showed that the bulk-driven LDO can achieve a significantly lower minimum supply voltage, i.e., 0.6 V, as compared to the gate-driven counterpart, 1 V, under the same reference voltage and load current conditions.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Capítulo o parte de un libro (BookPart)

CDS Free Frame Differencing Event Vision Pixel with Lateral Overflow Capacitor for Dynamic Range Extension

  • Jaklin, Marko
  • García Lesta, Daniel
  • López Martínez, Paula
  • Brea Sánchez, Víctor Manuel
This paper reports on a redesign of an event pixel by frame differencing with a lateral overflow capacitor for dynamic range extension. In the first prototype, the inclusion of an inpixel correlated double sampling (CDS) and an algorithm for dynamic range extension resulted in a large pixel area, with a 32μm pitch in 180 nm CMOS technology. This paper studies a CDS free version of our former pixel through post-layout simulations. The pixel pitch is shrinked to 22μm, power consumption is lowered and speed is preserved.
Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publicación Capítulo o parte de un libro (BookPart)

Live demonstration: a frame-based CMOS vision sensor with high dynamic range for events generation

  • Jaklin, Marko
  • García Lesta, Daniel
  • López Martínez, Paula
  • Brea Sánchez, Víctor Manuel
This live demonstration shows a frame-based CMOS vision sensor with high dynamic range for event generation. Our CMOS vision sensor features 64 × 64 processing elements that comprise one 4T-APS and local circuitry to provide events and high dynamic range extension. The event generation is performed synchronously through the threshold of the frame difference between consecutive frames. The dynamic range extension is carried out per-pixel with the overflow capacitance method. The sensor can reach up to thousands of event frames per second. Electrical simulations indicate a dynamic range of 85 dB, which narrows the gap with dynamic vision sensors