Artificial Intelligence‐Enabled Sensing Technologies in the 5G/Internet of Things Era: From Virtual Reality/Augmented Reality to the Digital Twin

Advanced Intelligent Systems, Journal Year: 2022, Volume and Issue: 4(7)

Published: March 29, 2022

With the development of 5G and Internet Things (IoT), era big data‐driven product design is booming. In addition, artificial intelligence (AI) also emerging evolving by recent breakthroughs in computing power software architectures. this regard, digital twin, analyzing various sensor data with help AI algorithms, has become a cutting‐edge technology that connects physical virtual worlds, which sensors are highly desirable to collect environmental information. However, although existing technologies, including cameras, microphones, inertial measurement units, etc., widely used as sensing elements for applications, high‐power consumption battery replacement them still problem. Triboelectric nanogenerators (TENGs) self‐powered supply feasible platform realizing self‐sustainable low‐power systems. Herein, progress on TENG‐based intelligent systems, is, wearable electronics, robot‐related smart homes, followed prospective future enabled fusion technology, focused on. Finally, how apply systems IoT discussed.

Language: Английский

---

In‐Sensor Computing: Materials, Devices, and Integration Technologies

Advanced Materials, Journal Year: 2022, Volume and Issue: 35(37)

Published: July 9, 2022

The number of sensor nodes in the Internet Things is growing rapidly, leading to a large volume data generated at sensory terminals. Frequent transfer between sensors and computing units causes severe limitations on system performance terms energy efficiency, speed, security. To efficiently process substantial amount data, novel computation paradigm that can integrate functions into networks should be developed. in-sensor reduces also decreases high complexity by processing locally. Here, hardware implementation device array levels discussed. physical mechanisms lead unique response characteristics their corresponding are illustrated. In particular, bioinspired enable functionalities neuromorphic computation. integration technology discussed perspective future development provided.

Language: Английский

---

Adaptive Extreme Edge Computing for Wearable Devices

Frontiers in Neuroscience, Journal Year: 2021, Volume and Issue: 15

Published: May 11, 2021

Wearable devices are a fast-growing technology with impact on personal healthcare for both society and economy. Due to the widespread of sensors in pervasive distributed networks, power consumption, processing speed, system adaptation vital future smart wearable devices. The visioning forecasting how bring computation edge have already begun, an aspiration provide adaptive extreme computing. Here, we holistic view hardware theoretical solutions toward that can guidance research this computing era. We propose various biologically plausible models continual learning neuromorphic technologies sensors. To envision concept, systematic outline which prospective low latency scenarios platforms expected. successively describe potential landscapes processors exploiting complementary metal-oxide semiconductors (CMOS) emerging memory (e.g., memristive devices). Furthermore, evaluate requirements within terms footprint, latency, data size. additionally investigate challenges beyond hardware, algorithms could impede enhancement

Language: Английский

---

In-memory computing with emerging memory devices: Status and outlook

APL Machine Learning, Journal Year: 2023, Volume and Issue: 1(1)

Published: Feb. 14, 2023

In-memory computing (IMC) has emerged as a new paradigm able to alleviate or suppress the memory bottleneck, which is major concern for energy efficiency and latency in modern digital computing. While IMC concept simple promising, details of its implementation cover broad range problems solutions, including various technologies, circuit topologies, programming/processing algorithms. This Perspective aims at providing an orientation map across wide topic IMC. First, technologies will be presented, both conventional complementary metal-oxide-semiconductor-based emerging resistive/memristive devices. Then, architectures considered, describing their aim application. Circuits include popular crosspoint arrays other more advanced structures, such closed-loop ternary content-addressable memory. The same might serve completely different applications, e.g., array can used accelerating matrix-vector multiplication forward propagation neural network outer product backpropagation training. algorithms properties enable diversification functions discussed. Finally, main challenges opportunities presented.

Language: Английский

---

Implementation of Convolutional Neural Networks in Memristor Crossbar Arrays with Binary Activation and Weight Quantization

ACS Applied Materials & Interfaces, Journal Year: 2024, Volume and Issue: 16(1), P. 1054 - 1065

Published: Jan. 1, 2024

We propose a hardware-friendly architecture of convolutional neural network using 32 × memristor crossbar array having an overshoot suppression layer. The gradual switching characteristics in both set and reset operations enable the implementation 3-bit multilevel operation whole that can be utilized as 16 kernels. Moreover, binary activation function mapped to read voltage ground is introduced evaluate result training with boundary 0.5 its estimated gradient. Additionally, we adopt fixed kernel method, where inputs are sequentially applied differential pair scheme, reducing unused cell waste. has robust against device state variations, neuron circuit experimentally demonstrated on customized breadboard. Thanks analogue device, accurate vector–matrix multiplication (VMM) by combining sequential weights obtained through tuning array. In addition, feature images extracted VMM during hardware inference 100 test samples classified, classification performance off-chip compared software results. Finally, results depending tolerance statistically verified several cycles.

Language: Английский

---

Neuromorphic computing at scale

Nature, Journal Year: 2025, Volume and Issue: 637(8047), P. 801 - 812

Published: Jan. 22, 2025

Language: Английский

---

Brain-inspired computing via memory device physics

APL Materials, Journal Year: 2021, Volume and Issue: 9(5)

Published: May 1, 2021

In our brain, information is exchanged among neurons in the form of spikes where both space (which neuron fires) and time (when contain relevant information. Every connected to other by synapses, which are continuously created, updated, stimulated enable processing learning. Realizing brain-like neuron/synapse network silicon would artificial autonomous agents capable learning, adaptation, interaction with environment. Toward this aim, conventional microelectronic technology, based on complementary metal–oxide–semiconductor transistors von Neumann computing architecture, does not provide desired energy efficiency scaling potential. A generation emerging memory devices, including resistive switching random access (RRAM) also known as memristor, can offer a wealth physics-enabled capabilities, multiplication, integration, potentiation, depression, time-decaying stimulation, suitable recreate some fundamental phenomena human brain silico. This work provides an overview about status most recent updates brain-inspired neuromorphic devices. After introducing RRAM device technologies, we discuss main functionalities integration fire, dendritic filtering, short- long-term synaptic plasticity. For each these functions, their proposed implementation terms materials, structure, characteristics. The rich physics, nano-scale tolerance stochastic variations, ability process situ make devices promising technology for future hardware intelligence.

Language: Английский

---

Effect of the Threshold Kinetics on the Filament Relaxation Behavior of Ag‐Based Diffusive Memristors

Advanced Functional Materials, Journal Year: 2021, Volume and Issue: 32(15)

Published: Dec. 22, 2021

Abstract Owing to their unique features such as thresholding and self‐relaxation behavior diffusive memristors built from volatile electrochemical metallization (v‐ECM) devices are drawing attention in emerging memories neuromorphic computing areas temporal coding. Unlike the switching kinetics of non‐volatile ECM cells, relaxation dynamics still under investigation. Comprehension identification underlying physical processes during utmost importance optimize modulate performance threshold devices. In this study, Ag/HfO 2 /Pt v‐ECM investigated. Depending on amplitude duration applied voltage pulses, filament analyzed a comprehensive approach. This enables different mechanisms rate‐limiting steps for formation and, consequently, simulate using model modified ECM. New insights gained combined study outline significance growth process its time. knowledge can be directly transferred into optimization operation conditions circuits.

Language: Английский

---

Ferroelectric-based synapses and neurons for neuromorphic computing

Neuromorphic Computing and Engineering, Journal Year: 2022, Volume and Issue: 2(1), P. 012002 - 012002

Published: Jan. 7, 2022

Abstract The shift towards a distributed computing paradigm, where multiple systems acquire and elaborate data in real-time, leads to challenges that must be met. In particular, it is becoming increasingly essential compute on the edge of network, close sensor collecting data. requirements system operating are very tight: power efficiency, low area occupation, fast response times, on-line learning. Brain-inspired architectures such as spiking neural networks (SNNs) use artificial neurons synapses simultaneously perform low-latency computation internal-state storage with consumption. Still, they mainly rely standard complementary metal-oxide-semiconductor (CMOS) technologies, making SNNs unfit meet aforementioned constraints. Recently, emerging technologies memristive devices have been investigated flank CMOS technology overcome systems’ memory this review, we will focus ferroelectric technology. Thanks its CMOS-compatible fabrication process extreme energy rapidly affirming themselves one most promising for neuromorphic computing. Therefore, discuss their role emulating synaptic behaviors an power-efficient way.

Language: Английский

---

HfO₂-based resistive switching memory devices for neuromorphic computing

Neuromorphic Computing and Engineering, Journal Year: 2022, Volume and Issue: 2(4), P. 042001 - 042001

Published: Sept. 7, 2022

Abstract HfO 2 -based resistive switching memory (RRAM) combines several outstanding properties, such as high scalability, fast speed, low power, compatibility with complementary metal-oxide-semiconductor technology, possible high-density or three-dimensional integration. Therefore, today, RRAMs have attracted a strong interest for applications in neuromorphic engineering, particular the development of artificial synapses neural networks. This review provides an overview structure, properties and RRAM computing. Both widely investigated nonvolatile devices pioneering works about volatile are reviewed. The device is first introduced, describing mechanisms associated to filamentary path defects oxygen vacancies. programming algorithms described high-precision multilevel operation, analog weight update synaptic exploiting resistance dynamics devices. Finally, presented, illustrating both networks supervised training multilevel, binary stochastic weights. Spiking then presented ranging from unsupervised spatio-temporal recognition. From this overview, appears mature technology broad range computing systems.

Language: Английский

---