Edge Platforms Research Articles

Terminal Security Configuration Optimization under Edge Computing

It is an important task of system security management to configure terminal security policy reasonably in edge computing (EC) to implement necessary security protection for terminals and EC system. Therefore, this paper analyses the security goals and requirements of terminal access for multi-service EC systems under the power grid, conducts terminal security risk assessment based on the AHP algorithm, and proposes an edge computing platform terminal security configuration optimization strategy, creatively using the OS-ELM algorithm. A framework for cloud training and edge-side online learning to meet the real-time and lightweight communication needs of edge platforms.

Fusion-Catalyzed Pruning for Optimizing Deep Learning on Intelligent Edge Devices

The increasing computational cost of deep neural network models limits the applicability of intelligent applications on resource-constrained edge devices. While a number of neural network pruning methods have been proposed to compress the models, prevailing approaches focus only on parametric operators (e.g., convolution), which may miss optimization opportunities. In this paper, we present a novel fusion-catalyzed pruning approach, called FuPruner, which simultaneously optimizes the parametric and non-parametric operators for accelerating neural networks. We introduce an aggressive fusion method to equivalently transform a model, which extends the optimization space of pruning and enables non-parametric operators to be pruned in a similar manner as parametric operators, and a dynamic filter pruning method is applied to decrease the computational cost of models while retaining the accuracy requirement. Moreover, FuPruner provides configurable optimization options for controlling fusion and pruning, allowing much more flexible performance-accuracy trade-offs to be made. Evaluation with state-of-the-art residual neural networks on five representative intelligent edge platforms, Jetson TX2, Jetson Nano, Edge TPU, NCS, and NCS2, demonstrates the effectiveness of our approach, which can accelerate the inference of models on CIFAR-10 and ImageNet datasets.

Siamese Networks for Few-Shot Learning on Edge Embedded Devices

Edge artificial intelligence hardware targets mainly inference networks that have been pretrained on massive datasets. The field of few-shot learning looks for methods that allow a network to produce high accuracy even when only a few samples of each class are available. Siamese networks can be used to tackle few-shot learning problems and are unique because they do not require retraining on the new samples of the new classes. Therefore they are suitable for edge hardware accelerators which often do not include on-chip training capabilities. This work describes improvements to a baseline Siamese network and benchmarking of the improved network on edge platforms. The modifications to the baseline network included adding multi-resolution kernels, a hybrid training process as well a different embedding similarity computation method. This network shows an average accuracy improvement of up to 22% across 4 datasets in a 5-way, 1-shot classification task. Benchmarking results using three edge computing platforms (NVIDIA Jetson Nano, Coral Edge TPU and a custom convolutional neural network accelerator) show that a Siamese classifier can run on these devices at reasonable frame rates for real-time performance, between 3 frames per second (FPS) on Jetson Nano and 60 FPS on the Edge TPU. By increasing the weight sparsity during training, the inference time of a network with 25% weight sparsity increases by 10 FPS but with only 1% drop in accuracy.

Enabling On-Device CNN Training by Self-Supervised Instance Filtering and Error Map Pruning

This work aims to enable on-device training of convolutional neural networks (CNNs) by reducing the computation cost at training time. CNN models are usually trained on high-performance computers and only the trained models are deployed to edge devices. But the statically trained model cannot adapt dynamically in a real environment and may result in low accuracy for new inputs. On-device training by learning from the real-world data after deployment can greatly improve accuracy. However, the high computation cost makes training prohibitive for resource-constrained devices. To tackle this problem, we explore the computational redundancies in training and reduce the computation cost by two complementary approaches: 1) self-supervised early instance filtering on data level and 2) error map pruning (EMP) on the algorithm level. The early instance filter selects important instances from the input stream to train the network and drops trivial ones. The EMP further prunes out insignificant computations when training with the selected instances. Extensive experiments show that the computation and energy cost is substantially reduced without any or with marginal accuracy loss. For example, when training ResNet-110 on CIFAR-10, we achieve 67.8% computation saving while preserving full accuracy and 75.1% computation saving with a marginal accuracy loss of 1.3%. When training LeNet on MNIST, we save 79% computation while boosting accuracy by 0.2%. Besides, practical energy saving is measured on edge platforms. We achieve 67.6% energy saving when training ResNet-110 on mobile GPU and 74.1% energy saving when training LeNet on MCU without any accuracy loss.

Data-Driven Optimization for Cooperative Edge Service Provisioning With Demand Uncertainty

Multiaccess edge computing (MEC) empowers service providers (SPs) to run applications on the shared edge platforms in close proximity to mobile users, enabling ultralow latency access to a wide variety of cloud services. However, how to decide the amount of edge computing resources to rent for mobile service provisioning poses great challenges as the service demand is unknown to SPs a priori and may vary across the geographically distributed edge sites spatially and temporally. The resource rental decision also significantly affects SPs’ deploying profits since it is critical for service deployment and workload assignment. This article investigates the service provisioning problem in a cooperative edge computing system under service demand uncertainty. We develop a holistic solution to make two-timescale decisions on edge resource rental and workload assignment to maximize SP’s deploying profits. Briefly, we exploit historical service demand traces at the edge sites to characterize the uncertainty in a data-driven manner and formulate the edge service provisioning problem into a two-stage risk-averse optimization. To solve the formulated problem without compromising the data privacy, we propose an algorithm integrating Benders decomposition (BD) and alternating direction method of multipliers (ADMMs), which enables each edge site to keep the historical traces locally and participate in the optimization process. Based on real-world data sets, extensive simulations are conducted to validate the efficacy of our scheme.

OpenRTiST: End-to-End Benchmarking for Edge Computing

The growth of edge computing depends on large-scale deployments of edge infrastructure. Benchmarking applications are needed to compare the performance across different edge deployments and against device-only and cloud-only implementations. In this article, we present OpenRTiST, an open-source application that is simultaneously compute-intensive, bandwidth-hungry, and latency-sensitive. It implements a form of augmented reality that lets you “see the world through the eyes of an artist.” We compare end-to-end application latency over varying network conditions and measure performance across a variety of edge platforms. OpenRTiST is designed to be easily deployed and has been used to showcase the benefits of edge computing.

An Overview of Platforms for Big Earth Observation Data Management and Analysis

In recent years, Earth observation (EO) satellites have generated big amounts of geospatial data that are freely available for society and researchers. This scenario brings challenges for traditional spatial data infrastructures (SDI) to properly store, process, disseminate and analyze these big data sets. To meet these demands, novel technologies have been proposed and developed, based on cloud computing and distributed systems, such as array database systems, MapReduce systems and web services to access and process big Earth observation data. Currently, these technologies have been integrated into cutting edge platforms in order to support a new generation of SDI for big Earth observation data. This paper presents an overview of seven platforms for big Earth observation data management and analysis—Google Earth Engine (GEE), Sentinel Hub, Open Data Cube (ODC), System for Earth Observation Data Access, Processing and Analysis for Land Monitoring (SEPAL), openEO, JEODPP, and pipsCloud. We also provide a comparison of these platforms according to criteria that represent capabilities of the EO community interest.

Design and Implementation of Acoustic Source Localization on a Low-Cost IoT Edge Platform

The implementation of algorithms for acoustic source localization on edge platforms for the Internet of Things (IoT) is gaining momentum. Applications based on acoustic monitoring can greatly benefit from efficient implementations of such algorithms, enabling novel services for smart homes and buildings or ambient-assisted living. In this context, this brief proposes extreme low-cost sound source localization system composed of two microphones and the low power microcontroller module ESP32. A Direction-Of-Arrival (DOA) algorithm has been implemented taking into account the specific features of this board, showing excellent performance despite the memory constraints imposed by the platform. We have also adapted off-the-shelf low-cost microphone boards to the input requirements of the ESP32 Analog-to-Digital Converter. The processing has been optimized by leveraging in parallel both cores of the microcontroller to capture and process the audio in real time. Our experiments expose that we can perform real-time localization, with a processing time below 3.3 ms.

ILog: An Intelligent Device for Automatic Food Intake Monitoring and Stress Detection in the IoMT

Not knowing when to stop eating or how much food is too much can lead to many health issues. In iLog, we propose a system which can not only monitor but also create awareness for the user of how much food is too much. iLog provides information on the emotional state of a person along with the classification of eating behaviors to Normal-Eating or Stress-Eating. Chronic stress, uncontrolled or unmonitored food consumption, and obesity are intricately connected, even involving certain neurological adaptations. We propose a deep learning model for edge computing platforms which can automatically detect, classify and quantify the objects from the plate of the user. Three different paradigms where the idea of iLog can be performed are explored in this research. Two different edge platforms have been implemented in iLog. The platforms include mobile, as it is widely used, and a single board computer which can easily be a part of network for executing experiments with iLog-Glasses being the main wearable. The iLog model has produced an overall accuracy of 98% with an average precision of 85.8%.

Deep Learning based Fruit Freshness Classification and Detection with CMOS Image sensors and Edge processors

CMOS Image sensors play a vital role in the exponentially growing field of Artificial Intelligence (AI). Applications like image classification, object detection and tracking are just some of the many problems now solved with the help of AI, and specifically deep learning. In this work, we target image classification to discern between six categories of fruits — fresh/ rotten apples, fresh/ rotten oranges, fresh/ rotten bananas. Using images captured from high speed CMOS sensors along with lightweight CNN architectures, we show the results on various edge platforms. Specifically, we show results using ON Semiconductor’s global-shutter based, 12MP, 90 frame per second image sensor (XGS-12), and ON Semiconductor’s 13 MP AR1335 image sensor feeding into MobileNetV2, implemented on NVIDIA Jetson platforms. In addition to using the data captured with these sensors, we utilize an open-source fruits dataset to increase the number of training images. For image classification, we train our model on approximately 30,000 RGB images from the six categories of fruits. The model achieves an accuracy of 97% on edge platforms using ON Semiconductor’s 13 MP camera with AR1335 sensor. In addition to the image classification model, work is currently in progress to improve the accuracy of object detection using SSD and SSDLite with MobileNetV2 as the feature extractor. In this paper, we show preliminary results on the object detection model for the same six categories of fruits.

The Jet-I-U: A biomimetic assistive robot for the visually impaired and elderly

As the world’s population continues to grow in size and in age, it remains vital to develop streamlined and affordable technologies to address the needs of those who require companionship and care. The elderly and visually impaired remain a significant and ever-increasing segment of society, yet due to the cost of at-home health care, many of these individuals are unable to afford full-time caretakers to assist with everyday tasks. Although the luxury of at-home care remains elusive to many due to the associated cost, social robots are an important and meaningful way to overcome this issue. This research explores how combining the principles of biomimicry and social robotics with the Internet of Things (IoT) and artificial intelligence (AI) can enhance assistive technology. This paper proposes a novel assistive device called the Jet-I-U, which aims to support visually impaired and elderly individuals with day-to-day tasks. The Jet-I-U mimics the behavioral and physical characteristics of a pet scarlet macaw and provides both information and companionship to this demographic. The device emulates a scarlet macaw’s morphology and movement through laser-cut features and various biomimetic components. The intelligence and behavioral characteristics of the macaw are modeled by the application of IoT Edge platforms, AI, and Machine Learning (ML). By utilizing powerful off-the-shelf microcontrollers like the NVIDIA Jetson Nano and Raspberry Pi, this device is both affordable and practical. Experimental analysis was performed to evaluate Jet-I-U’s performance based on several criteria. The results demonstrated the versatility and applicability of the proposed solution as a companion to the visually impaired and elderly.

An efficient convolutional neural network for small traffic sign detection

Deep learning has become a ubiquitous method in object detection among multiple domains recently. However, in the era of edge computing, deploying deep neural networks on mobile edge platforms are challenging due to long latency and huge computational cost. As previous research efforts were usually focused on accuracy, achieving the balance between computational consumption and accuracy is a more significant problem to be tackled in mobile edge computing domain. To this end, we proposed an efficient convolutional neural network (CNN), which can remarkably minimize the redundancy, reduce the parameters and speed up the networks. The effectiveness of the network is further proved with experiments on a Tsinghua-Tencent 100K traffic sign dataset. Results show that under the same-level model size, our network outperforms the state-of-the-art Fast R-CNN and Faster R-CNN with 10% improvement in accuracy. Compared to similar work, the computational consumption on running time and memory of our network has been also reduced in the premise of little loss in accuracy.

Multilayer Architecture Model for Mobile Cloud Computing Paradigm

Mobile Cloud Computing is one of today′s more disruptive paradigms of computation due to its effects on the performance of mobile computing and the development of Internet of Things. It is able to enhance the capabilities of devices by outsourcing the workload to external computing platforms deployed along the network, such as cloud servers, cloudlets, or other edge platforms. The research described in this work presents a computational model of a multilayer architecture for increasing the performance of devices using the Mobile Cloud Computing paradigm. The main novelty of this work lies in defining a comprehensive model where all the available computing platforms along the network layers are involved to perform the outsourcing of the application workload. This proposal provides a generalization of the Mobile Cloud Computing paradigm which allows handling the complexity of scheduling tasks in such complex scenarios. The behaviour of the model and its ability of generalization of the paradigm are exemplified through simulations. The results show higher flexibility for making offloading decisions.

Transformation to Advanced Mechatronics Systems Within New Industrial Revolution: A Novel Framework in Automation of Everything (AoE)

The recent advances in cyber-physical domains, cloud, cloudlet, and edge platforms along with the evolving Artificial Intelligence (AI) techniques, big data analytics, and cutting-edge wireless communication technologies within the Industry 4.0 (4IR) are urging mechatronics designers, practitioners, and educators to further review the ways in which mechatronics systems are perceived, designed, manufactured, and advanced. Within this scope, we introduce the service-oriented cyber-physical advanced mechatronics systems (AMSs) along with current and future challenges. The objective in AMSs is to create remarkably intelligent autonomous products by 1) forging effective sensing, self-learning, Wisdom as a Service (WaaS), Information as a Service (InaaS), precise decision making, and actuation using effective location-independent monitoring, control and management techniques with products and 2) maintaining a competitive edge through better product performances via immediate and continuous learning, while the products are being used by customers and are being produced in factories within the cycle of Automation of Everything (AoE). With the advanced wireless communication techniques and improved battery technologies, the AMSs are capable of getting independent and working with other massive AMSs to construct robust, customizable, energy-efficient, autonomous, intelligent, and immersive platforms. In this regard, rather than providing technological details, this paper implements philosophical insights into 1) how mechatronics systems are being transformed into AMSs; 2) how robust AMSs can be developed by both exploiting the wisdom created within cyber-physical smart domains in the edge and cloud platforms and incorporating all the stakeholders with diverse objectives into all phases of the product life-cycle; and 3) what essential common features AMSs should acquire to increase the efficacy of products and prolong their product life. Against this background, an AMS development framework is proposed in order to contextualize all the necessary phases of AMS development and direct all stakeholders to rivet high-quality products and services within AoE.

A two-stage locality-sensitive hashing based approach for privacy-preserving mobile service recommendation in cross-platform edge environment

With the increasing popularity of service computing paradigm, tremendous resources or services are emerging rapidly on the Web, imposing heavy burdens on the service selection decisions of users. In this situation, recommendation (e.g., collaborative filtering) has been considered as one of the most effective ways to alleviate such burdens. However, in the mobile and edge environment, the service recommendation bases, i.e., historical service usage data are often generated from various mobile devices (e.g., Smartphone and PDA) and stored in different edge platforms. Therefore, effective collaboration between these distributed edge platforms plays an important role in the successful mobile service recommendation. Such a cross-platform collaboration process often faces the following two challenges. First, a platform is often reluctant to release its data to other platforms due to privacy concerns. Second, the collaboration efficiency is often low when the data in each platform update frequently. In view of these two challenges, we introduce MinHash, an instance of Locality-Sensitive Hashing (LSH), into service recommendation, and further put forward a novel privacy-preserving and scalable mobile service recommendation approach based on two-stage LSH, namedSerRectwo-LSH. Finally, extensive experiments are conducted on WS-DREAM, a real distributed service quality dataset, and the evaluation results demonstrate that both the service recommendation accuracy and the scalability have been significantly improved while privacy preservation is guaranteed.

Privacy‐Aware Multidimensional Mobile Service Quality Prediction and Recommendation in Distributed Fog Environment

With the ever‐increasing popularity of mobile computing technology, a wide range of computational resources or services (e.g., movies, food, and places of interest) are migrating to the mobile infrastructure or devices (e.g., mobile phones, PDA, and smart watches), imposing heavy burdens on the service selection decisions of users. In this situation, service recommendation has become one of the promising ways to alleviate such burdens. In general, the service usage data used to make service recommendation are produced by various mobile devices and collected by distributed edge platforms, which leads to potential leakage of user privacy during the subsequent cross‐platform data collaboration and service recommendation process. Locality‐Sensitive Hashing (LSH) technique has recently been introduced to realize the privacy‐preserving distributed service recommendation. However, existing LSH‐based recommendation approaches often consider only one quality dimension of services, without considering the multidimensional recommendation scenarios that are more complex but more common. In view of this drawback, we improve the traditional LSH and put forward a novel LSH‐based service recommendation approach named SerRecmulti-qos, to protect users’ privacy over multiple quality dimensions during the distributed mobile service recommendation process.

Virtual Resources & Blockchain for Configuration Management in IoT

Until now, most systems for Internet of Things (IoT) management, have been designed in a Cloud-centric manner, getting benefits from the unified platform that the Cloud offers. However, a Cloud-centric infrastructure mainly achieves static sensor and data streaming systems, which do not support the direct configuration management of IoT components. To address this issue, a virtualization of IoT components (Virtual Resources) is introduced at the edge of the IoT network. This research also introduces permission-based Blockchain protocols to handle the provisioning of Virtual Resources directly onto edge devices. The architecture presented by this research focuses on the use of Virtual Resources and Blockchain protocols as management tools to distribute configuration tasks towards the edge of the IoT network. Results from lab experiments demonstrate the successful deployment and communication performance (response time in milliseconds) of Virtual Resources on two edge platforms, Raspberry Pi and Edison board. This work also provides performance evaluations of two permission-based blockchain protocol approaches. The first blockchain approach is a Blockchain as a Service (BaaS) in the Cloud, Bluemix. The second blockchain approach is a private cluster hosted in a Fog network, Multichain

POINT-OF-NEED DIAGNOSTICS: BIOSURVEILLANCE WITH A DEVICE2CLOUD CAPABILITY IN SIERRA LEONE

BackgroundInfectious diseases contribute to a high burden of diseases globally. Surveillance using low-cost technology, combined with cutting edge platforms offers a path for understanding the disease ecology of locations of...

On brain activity mapping: insights and lessons from Brain Decoding Project to map memory patterns in the hippocampus

The BRAIN project recently announced by the president Obama is the reflection of unrelenting human quest for cracking the brain code, the patterns of neuronal activity that define who we are and what we are. While the Brain Activity Mapping proposal has rightly emphasized on the need to develop new technologies for measuring every spike from every neuron, it might be helpful to consider both the theoretical and experimental aspects that would accelerate our search for the organizing principles of the brain code. Here we share several insights and lessons from the similar proposal, namely, Brain Decoding Project that we initiated since 2007. We provide a specific example in our initial mapping of real-time memory traces from one part of the memory circuit, namely, the CA1 region of the mouse hippocampus. We show how innovative behavioral tasks and appropriate mathematical analyses of large datasets can play equally, if not more, important roles in uncovering the specific-to-general feature-coding cell assembly mechanism by which episodic memory, semantic knowledge, and imagination are generated and organized. Our own experiences suggest that the bottleneck of the Brain Project is not only at merely developing additional new technologies, but also the lack of efficient avenues to disseminate cutting edge platforms and decoding expertise to neuroscience community. Therefore, we propose that in order to harness unique insights and extensive knowledge from various investigators working in diverse neuroscience subfields, ranging from perception and emotion to memory and social behaviors, the BRAIN project should create a set of International and National Brain Decoding Centers at which cutting-edge recording technologies and expertise on analyzing large datasets analyses can be made readily available to the entire community of neuroscientists who can apply and schedule to perform cutting-edge research.