Share This Special Issue

Special Issue Editors

Special Issue Information

Dear Colleagues,

Edge computing is becoming a widely adopted technological paradigm. It is especially valid in application domains such as smart industry, smart city, smart ports, and, more generally, in the design of robust and reliable applications where bandwidth, privacy, latency, and responsiveness impose strong constraints. The original idea of edge computing comes from pushing the execution of cloud-scale functionalities at the edge of the network, close to where data are produced. Edge computing has required hard re-engineering efforts, especially concerning how applications are designed, developed, packaged, and deployed, in consideration of the new challenges imposed by different and usually constrained execution environments (e.g., embedded PCs, PLCs, single-board computers, microcontrollers). At the same time, the hype surrounding artificial intelligence (AI) has strongly influenced how its models and tools are conceived or adapted to fit the small capacities of far edge/IoT devices. Around this fulcrum, two new exciting research and innovation directions are gaining momentum, namely edgeAI and TinyML, unleashing new challenges regarding how data are collected and processed, and how AI model architectures are designed, adapted, optimized, deployed, updated, trained, and executed along the cloud-to-thing continuum.

Topics relevant to this Special Issue include, but are not limited to, the following:

EdgeAI and TinyML in practical applications;
Edge computing architectures supporting AI deployment;
Methods for the optimization of models for AI at the edge/far-edge of the network;
Methods to adapt cloud-scale AI models for the edge;
Training of AI models at the edge;
Transfer learning of AI model at the edge of the network;
Novel training approaches for AI models at the edge;
Distributed and/or decentralized orchestration of AI pipelines along the cloud-to-thing continuum;
Federated learning approaches with tiny IoT devices.

Prof. Dr. Massimo Vecchio
Dr. Mattia Antonini
Dr. Miguel Pincheira
Dr. Panagiotis Trakadas
Dr. Paolo Azzoni
Guest Editors

Manuscript Submission Information

Manuscripts should be submitted online at www.mdpi.com by registering and logging in to this website. Once you are registered, click here to go to the submission form. Manuscripts can be submitted until the deadline. All submissions that pass pre-check are peer-reviewed. Accepted papers will be published continuously in the journal (as soon as accepted) and will be listed together on the special issue website. Research articles, review articles as well as short communications are invited. For planned papers, a title and short abstract (about 100 words) can be sent to the Editorial Office for announcement on this website.

Submitted manuscripts should not have been published previously, nor be under consideration for publication elsewhere (except conference proceedings papers). All manuscripts are thoroughly refereed through a single-blind peer-review process. A guide for authors and other relevant information for submission of manuscripts is available on the Instructions for Authors page. Sensors is an international peer-reviewed open access semimonthly journal published by MDPI.

Please visit the Instructions for Authors page before submitting a manuscript. The Article Processing Charge (APC) for publication in this open access journal is 2600 CHF (Swiss Francs). Submitted papers should be well formatted and use good English. Authors may use MDPI's English editing service prior to publication or during author revisions.

Keywords

Internet of Things
embedded intelligence
deep learning
MLOps
TinyMLOps
self-supervised learning
far edge
cloud-to-thing continuum

Published Papers (1 paper)

Order results

Result details

Show export options Show export options

Select all

Export citation of selected articles as:

Research

17 pages, 468 KiB

Open AccessArticle

AoCStream: All-on-Chip CNN Accelerator with Stream-Based Line-Buffer Architecture and Accelerator-Aware Pruning

by Hyeong-Ju Kang and Byung-Do Yang

Sensors 2023, 23(19), 8104; https://doi.org/10.3390/s23198104 - 27 Sep 2023

Viewed by 702

Abstract

Convolutional neural networks (CNNs) play a crucial role in many EdgeAI and TinyML applications, but their implementation usually requires external memory, which degrades the feasibility of such resource-hungry environments. To solve this problem, this paper proposes memory-reduction methods at the algorithm and architecture level, implementing a reasonable-performance CNN with the on-chip memory of a practical device. At the algorithm level, accelerator-aware pruning is adopted to reduce the weight memory amount. For activation memory reduction, a stream-based line-buffer architecture is proposed. In the proposed architecture, each layer is implemented by a dedicated block, and the layer blocks operate in a pipelined way. Each block has a line buffer to store a few rows of input data instead of a frame buffer to store the whole feature map, reducing intermediate data-storage size. The experimental results show that the object-detection CNNs of MobileNetV1/V2 and an SSDLite variant, widely used in TinyML applications, can be implemented even on a low-end FPGA without external memory. Full article

(This article belongs to the Special Issue The Rise of EdgeAI and TinyML for the Next-Generation IoT)

► Show Figures

Journal Menu

Journal Browser

The Rise of EdgeAI and TinyML for the Next-Generation IoT

Share This Special Issue

Special Issue Editors

Special Issue Information

Keywords

Published Papers (1 paper)

Research

Further Information

Guidelines

MDPI Initiatives

Follow MDPI