Low Computational Resources Research Articles

IoT Edge Caching: Taxonomy, Use Cases and Perspectives

Low latency is an essential constraint for many emerging Internet of Things (IoT) applications. Cloud-based solutions can impose excessive network and processing latency that might hinder the system's operation. On the other hand, offloading tasks to the edge layer - close to where data is consumed - might be unfeasible in many situations due to the low computational resources of edge devices or because the system utilizes third-party applications. One potential solution is caching data on the edge of the network, thus decreasing the latency of subsequent requests, saving bandwidth, and avoiding redundant computation. At the same time, caching IoT data on the edge has its challenges and features that deviate from the standard methods of network caching, such as the content dynamicity and the power constraints. This paper performs a comprehensive review of the state-of-art in IoT Edge Caching and proposes a novel taxonomy focused on five orthogonal features of edge caching: placement, distance, strategy, metrics, and design. Moreover, we illustrate the characteristics identified in five prevalent use cases for the IoT. Finally, we conduct an in-depth analysis and performance evaluation of edge caching deployed in a real Structural Health Monitoring (SHM) scenario. Results show that different caching strategies decrease the system latency by more than 95 percent.

HHI-AttentionNet: An Enhanced Human-Human Interaction Recognition Method Based on a Lightweight Deep Learning Model with Attention Network from CSI.

Nowadays WiFi based human activity recognition (WiFi-HAR) has gained much attraction in an indoor environment due to its various benefits, including privacy and security, device free sensing, and cost-effectiveness. Recognition of human-human interactions (HHIs) using channel state information (CSI) signals is still challenging. Although some deep learning (DL) based architectures have been proposed in this regard, most of them suffer from limited recognition accuracy and are unable to support low computation resource devices due to having a large number of model parameters. To address these issues, we propose a dynamic method using a lightweight DL model (HHI-AttentionNet) to automatically recognize HHIs, which significantly reduces the parameters with increased recognition accuracy. In addition, we present an Antenna-Frame-Subcarrier Attention Mechanism (AFSAM) in our model that enhances the representational capability to recognize HHIs correctly. As a result, the HHI-AttentionNet model focuses on the most significant features, ignoring the irrelevant features, and reduces the impact of the complexity on the CSI signal. We evaluated the performance of the proposed HHI-AttentionNet model on a publicly available CSI-based HHI dataset collected from 40 individual pairs of subjects who performed 13 different HHIs. Its performance is also compared with other existing methods. These proved that the HHI-AttentionNet is the best model providing an average accuracy, F1 score, Cohen’s Kappa, and Matthews correlation coefficient of 95.47%, 95.45%, 0.951%, and 0.950%, respectively, for recognition of 13 HHIs. It outperforms the best existing model’s accuracy by more than 4%.

Vision-Based Robotic Grasping of Reels for Automatic Packaging Machines

In this work, we present a vision system particularly suited to the automatic recognition of reels in the field of automatic packaging machines. The output of the vision system is used to guide the autonomous grasping of the reels by a robot for a subsequent manipulation task. The proposed solution is built around three different methods to solve the ellipse-detection problem in an image. Such methods leverage standard image processing and mathematical algorithms, which are tailored to the targeted application. An experimental campaign demonstrates the efficacy of the proposed approach, even in the presence of low computational power and limited hardware resources, as in the use-case at hand.

Class incremental learning for wildlife biodiversity monitoring in camera trap images

Camera traps have been widely used for wildlife biodiversity monitoring, providing abundant ecological data. Manually classifying such abundant images is time-consuming and labor-intensive. Existing deep learning methods solve this problem for a fixed set of predefined wildlife species. The model trained on such sets cannot be applied to new wildlife species. Retraining models on new wildlife species can lead to catastrophic forgetting. Thus, in this work, we propose a class incremental learning method to identify new wildlife species. Our method employs a novel adaptive exemplar assignment (AEA) strategy with dynamic exemplar amounts to adapt to new species while alleviating the forgetting of old ones. Due to memory constraints, the data imbalance between limited exemplars and new species data can lead to class bias. We mitigate it by performing center vector retrieval (CVR) to classify samples in feature space and bypass the biased linear classifier. In addition, we propose two variants of CVR that incorporate the advantage of the linear classifier to further improve the performance. By using only 4% of old species data, our method achieves 77.09% accuracy at a low computational resource for recognition. Through extensive experiments and ablations, we demonstrate the superiority of our proposed approach over state-of-the-art methods. This method facilitates wildlife monitoring, biodiversity conservation, and ecological assessment.

A New Software-Based Optimization Technique for Embedded Latency Improvement of a Constrained MIMO MPC

Embedded controllers for multivariable processes have become a powerful tool in industrial implementations. Here, the Model Predictive Control offers higher performances than standard control methods. However, they face low computational resources, which reduces their processing capabilities. Based on pipelining concept, this paper presents a new embedded software-based implementation for a constrained Multi-Input-Multi-Output predictive control algorithm. The main goal of this work focuses on improving the timing performance and the resource usage of the control algorithm. Therefore, a profiling study of the baseline algorithm is developed, and the performance bottlenecks are identified. The functionality and effectiveness of the proposed implementation are validated in the NI myRIO 1900 platform using the simulation of a jet transport aircraft during cruise flight and a tape transport system. Numerical results for the study cases show that the latency and the processor usage are substantially reduced compared with the baseline algorithm, 4.6× and 3.17× respectively. Thus, efficient program execution is obtained which makes the proposed software-based implementation mainly suitable for embedded control systems.

IoTRepair: Flexible Fault Handling in Diverse IoT Deployments

IoT devices can be used to complete a wide array of physical tasks, but due to factors such as low computational resources and distributed physical deployment, they are susceptible to a wide array of faulty behaviors. Many devices deployed in homes, vehicles, industrial sites, and hospitals carry a great risk of damage to property, harm to a person, or breach of security if they behave faultily. We propose a general fault handling system named IoTRepair, which shows promising results for effectiveness with limited latency and power overhead in an IoT environment. IoTRepair dynamically organizes and customizes fault-handling techniques to address the unique problems associated with heterogeneous IoT deployments. We evaluate IoTRepair by creating a physical implementation mirroring a typical home environment to motivate the effectiveness of this system. Our evaluation showed that each of our fault-handling functions could be completed within 100 milliseconds after fault identification, which is a fraction of the time that state-of-the-art fault-identification methods take (measured in minutes). The power overhead is equally small, with the computation and device action consuming less than 30 milliwatts. This evaluation shows that IoTRepair not only can be deployed in a physical system, but offers significant benefits at a low overhead.

Tied Bilateral learning for Aquaculture Image Enhancement

Aquaculture plays a critical role in food security and nutrition strategies. The application of intelligent aquaculture technology has shown promising performance in improving aquaculture productivity and increasing economic benefits with its rapid advancement and good prospects. However, degraded underwater images have hampered the existing computer vision applications in intelligent aquaculture. To this end, a novel Tied Bilateral learning network is proposed for Aquaculture Image Enhancement (TBAIE), which improves the degraded aquaculture images to meet the requirements of various computer vision applications in aquaculture. Concretely, a novel multiple tied guidance module is designed to generate a multi-channel feature map and capture long-range features based on input. Then, a feature fusion module is introduced with a novel tied attention block to blend the feature and suppress noise with a low computational resource. Experimental results demonstrate that the proposed TBAIE can improve the quality of aquaculture images and remove color distortion. Moreover, TBAIE can achieve state-of-the-art in quantitative and qualitative metrics and meet the practical requirements of different aquaculture vision tasks.

CE-text: A context-Aware and embedded text detector in natural scene images

• A novel deep and context-aware CNN structure for accurate and fast text detection • Hierarchically channel wise attention scheme with channel wise and multilayer features • Adopts frequency-based deep compression method to build a lightweight text detector With the significant power of deep learning architectures, researchers have made much progress on effectiveness and efficiency of text detection in the past few years. However, due to the lack of consideration of unique characteristics of text components, directly applying deep learning models to perform text detection task is prone to result in low accuracy, especially producing false positive detection results. To ease this problem, we propose a lightweight and context-aware deep convolutional neural network (CNN) named as CE-Text, which appropriately encodes multi-level channel attention information to construct discriminative feature map for accurate and efficient text detection. To fit with low computation resource of embedded systems, we further transform CE-Text into a lighter version with a frequency based deep CNN compression method, which expands applicable scenarios of CE-Text into variant embedded systems. Experiments on several popular datasets show that CE-Text not only has achieved accurate text detection results in scene images, but also could run with fast performance in embedded systems.

Semantic Self-Segmentation for Abstractive Summarization of Long Documents in Low-Resource Regimes

The quadratic memory complexity of transformers prevents long document summarization in low computational resource scenarios. State-of-the-art models need to apply input truncation, thus discarding and ignoring potential summary-relevant contents, leading to a performance drop. Furthermore, this loss is generally destructive for semantic text analytics in high-impact domains such as the legal one. In this paper, we propose a novel semantic self-segmentation (Se3) approach for long document summarization to address the critical problems of low-resource regimes, namely to process inputs longer than the GPU memory capacity and produce accurate summaries despite the availability of only a few dozens of training instances. Se3 segments a long input into semantically coherent chunks, allowing transformers to summarize very long documents without truncation by summarizing each chunk and concatenating the results. Experimental outcomes show the approach significantly improves the performance of abstractive summarization transformers, even with just a dozen of labeled data, achieving new state-of-the-art results on two legal datasets of different domains and contents. Finally, we report ablation studies to evaluate each contribution of the components of our method to the performance gain.

Comparative study of constitutive relations implemented in RELAP5 and TRACE – Part I: Methodology & wall friction

Nuclear thermal-hydraulic system analysis codes have been developed to simulate nuclear reactor systems, which solve simplified governing equations by replacing source terms with constitutive relations for simulating entire reactor systems with low computational resources. For half a century, many efforts have been made for wider versatility and higher accuracy of system codes, but various factors can affect the code analysis results, and it was difficult to isolate these factors and interpret them individually. In this study, two system codes, RELAP5 and TRACE, which have many users and are highly reliable, are selected to analyze only the effects of constitutive relations. The influence of constitutive relations is analyzed using in-house platforms that replicate constitute relations of RELAP5 and TRACE equally to exclude factors that may affect analysis results, such as governing equation solvers and user effects. Among the various constitutive relations, the analysis is performed on the wall variables expected to have the most influence on the analysis results. Part 1 paper presents the methodology and wall friction model comparison, while Part 2 paper shows wall heat transfer comparison of the two selected codes.

FOCUSeR: A Fog Online Context-Aware Up-to-Date Sensor Ranking Method

Data obtained from sensors connected to wireless sensor networks must be stored and processed to enable environments such as smart cities. However, with the exponential growth in the number of devices at the edge of the network, it is necessary to implement robust techniques, capable of selecting reliable data sources and meeting low latency requirements, in order to serve critical applications. Thus, to overcome these challenges, this research work presents FOCUSeR, a method for ranking sensors. The method uses the evaluation of data as a criterion for the ranking, allowing us to identify occurrences of failures in sensors and anomalies in environments. In order to meet the requirements inherent to WSNs, the proposed method was developed to run in a fog computing environment, using online learning and constant updating over time to avoid effects such as time drift. The generated ranking lists are managed through distributed hash tables. To provide reliability to the experimental results, a real experimental environment was developed. Moreover, using this developed testbed, a dataset with labels was created, to support the evaluation of the method. In addition, four other real datasets were used, three of which were labeled through artificial fault injection. These datasets were labeled in a related work that focused on injecting artificial faults. The experimental results obtained demonstrate that the proposed approach can provide reliability in the use of sensor data, using low computational resources and reducing latency in the sensor selection process. Precision rates are approximately 98% and Accuracy rates are greater than 94% across all datasets. In addition, the analyses carried out show that the Accuracy has an increasing rate as the number of samples also increases. Results obtained in the failure data recovery also demonstrate the feasibility of the proposal in this resource.

ANALYZING ARCHITECTURES FOR NEURAL MACHINE TRANSLATION USING LOW COMPUTATIONAL RESOURCES

With the recent developments in the field of Natural Language Processing, there has been a rise in the use of different architectures for Neural Machine Translation. Transformer architectures are used to achieve state-of-the-art accuracy, but they are very computationally expensive to train. Everyone cannot have such setups consisting of high-end GPUs and other resources. We train our models on low computational resources and investigate the results. As expected, transformers outperformed other architectures, but there were some surprising results. Transformers consisting of more encoders and decoders took more time to train but had fewer BLEU scores. LSTM performed well in the experiment and took comparatively less time to train than transformers, making it suitable to use in situations having time constraints.

Mesoscale simulation of typhoon-generated storm surge: methodology and Shanghai case study

Abstract. The increasing vulnerability of coastal megacities to storm surge inundation means both infrastructure and populations are subject to significant threat. Planning for further urban development should include consideration of the changing circumstances in coastal cities to ensure a sustainable future. A sustainable urban plan relies on sound preparedness and prediction of future climate change and multiple natural hazards. In light of these needs for urban planning, this paper develops a general method to simulate typhoon-generated storm surge at the mesoscale (1–100 km in length). Mesoscale simulation provides a general approach with reasonable accuracy that could be implemented for planning purposes while having a relatively low computation resource requirement. The case study of Shanghai was chosen to implement this method. The mesoscale simulations of two historical typhoons not only provides realistic typhoon storm surge inundation results at the city level but is also suitable for implementing a large amount of simulations for future scenario studies. The method will be generally applicable to all coastal cities around the world to examine the effect of future climate change on typhoon-generated storm surge even when historical observation data are inadequate or not available.

MODELING OF THE FLOW AROUND A SQUARE CYLINDER WITH HYBRID TURBULENCE MODELS

The development of computing resources has played an exceptional role in the development of turbulence modeling. However, simulation of a real flow is still a significant problem if high accuracy of simulation results with low computational resources is required. Therefore, the use of hybrid models is a promising direction for solving this problem of computational fluid dynamics. When using hybrid models, modeling of near-wall flows is carried out using RANS models, while far from the wall, LES models are used. In this work, we used the k - ω SST and DDES models. The aim of the work is to identify a suitable turbulence model that could give accurate and reliable results for practical engineering applications. To evaluate the model used in this work, the problem of modeling the air flow around a square cylinder is considered. Comparison of the results of numerical calculations with the corresponding experimental data shows that the use of the hybrid k - ω SST()-DDES model is promising.

3-D Metamaterials: Trends on Applied Designs, Computational Methods and Fabrication Techniques

Metamaterials are artificially engineered devices that go beyond the properties of conventional materials in nature. Metamaterials allow for the creation of negative refractive indexes; light trapping with epsilon-near-zero compounds; bandgap selection; superconductivity phenomena; non-Hermitian responses; and more generally, manipulation of the propagation of electromagnetic and acoustic waves. In the past, low computational resources and the lack of proper manufacturing techniques have limited attention towards 1-D and 2-D metamaterials. However, the true potential of metamaterials is ultimately reached in 3-D configurations, when the degrees of freedom associated with the propagating direction are fully exploited in design. This is expected to lead to a new era in the field of metamaterials, from which future high-speed and low-latency communication networks can benefit. Here, a comprehensive overview of the past, present, and future trends related to 3-D metamaterial devices is presented, focusing on efficient computational methods, innovative designs, and functional manufacturing techniques.

Disjoint multipath RPL for QoE/QoS provision in the Internet of Multimedia Things

RPL is the IETF standardized IPv6 Routing Protocol for Low-power and lossy networks (LLNs). The major effort is therefore made on handling low data rate traffic. Meanwhile, the Internet of Multimedia Things emerged as one hot topic of the Internet of Things. In LLNs, providing a descent quality of service (QoS) along with a user quality of experience (QoE) for multimedia applications is challenging. High bandwidth and significant computation capabilities are necessary while LLNs are very resource constrained. To raise the available bandwidth to accommodate high data rate applications, we propose to simultaneously transmit a flow on multiple disjoint paths. To do so, we exploit the DODAG structure, already maintained by RPL, to build disjoint paths without incurring extra overhead. To reduce the amount of data to send while considering the low computation resources of video sensors, we propose to apply a priority-based low-complexity encoding scheme on the captured pictures. Highest priority data can be replicated on more than one path. Our conducted experiments using simulations as well as a real testbed show that the best QoS and QoE are obtained when multiple paths are used along with the replication of highest priority data packets.

A Compact Representation of Indoor Trajectories

In this article, we present a system that combines indoor positioning with a compression algorithm for trajectories in the context of a nursing home. Our aim is to gather and effectively represent the location of the residents and caregivers along time, while allowing for efficient access to those data. We briefly show the system architecture that enables the automatic tracking of user’s movements and consequently the gathering of their locations. Then, we present indRep, our compact representation to handle positioning data using grammar-based compression, and provide two basic operations that enable pseudorandom access to the data. Finally, we include experiments that show that indRep is competitive with well-known general-purpose compressors in terms of compression effectiveness and also provides fast access to the compressed data. We expect both features would enable exploitation functionalities even in computers with rather low computational resources.

A Machine Learning Pipeline for Measurement of Arterial Stiffness in A-Mode Ultrasound.

Arterial stiffness (AS) of the carotid artery is an early marker of stratifying cardiovascular disease risk. This article aims to improve the performance of ARTSENS, a noninvasive A-mode ultrasound-based device for measuring AS. The primary objective of ARTSENS is to enable the measurement of elastic modulus using A-Mode ultrasound and blood pressure. As this device is image-free, there is a need to automate: 1) carotid detection; 2) wall localization; and 3) inner lumen diameter measurement. This has been performed using conventional signal processing methods in some of the earlier works in this domain. In this article, deep neural network (DNN) models are employed to perform the above three tasks. The DNNs were trained over data acquired from 82 subjects at two different medical centers. Ground-truth labeling was performed by a trained operator using corresponding measurements from the state-of-the-art Aloka e-Tracking system. All three DNN models had significantly lower errors compared to earlier signal processing methods and could perform their measurements using a single A-Mode frame. Using the DNNs, two different machine learning pipelines have been proposed here to measure the elastic modulus; the best among them could achieve an error of 9.3% with the Pearson correlation coefficient of 0.94 ( ). The models were tested on Raspberry Pi and Jetson Nano single board computers to demonstrate real-time processing on low computational resources.

Power System Frequency and Phasor Estimation for a Low-Cost Synchrophasor Device Using the Nonlinear Least-Square Method

Wide area monitoring, protection, and control has now become an essential infrastructure to run an interconnected power system. To accomplish this by using synchrophasor technology, a wide range of phasor measurement unit (PMU) designs are available in the literature, but most of them are very costly because of the requirement of a high-speed processor to run complex algorithms. In this article, an algorithm for the synchrophasor and frequency estimation is proposed for processing platforms with low computational resources and is delineated extensively in respect of both accuracy and processing time. The proposed solution harnesses the main advantages of two well-known algorithms, viz. the zero-crossing detection and nonlinear least-squares method. Both algorithms are incorporated in a computationally efficient manner to reduce the processing time up to a lower extent while maintaining a good accuracy as per the standard IEEE testing conditions. The performance of the proposed algorithm has been evaluated through simulations as well as experimentally on the Raspberry Pi board. The consistency of the test results and minimum processing time provides clear evidence that the implemented design is suitable for PMU prototyping and is compliant with the mandatory IEEE standard C37.118.1-2011 and its amendment IEEE C37.118.1a-2014.

MD Loss: Efficient Training of 3-D Seismic Fault Segmentation Network Under Sparse Labels by Weakening Anomaly Annotation

Data-driven fault detection has been regarded as a 3D image segmentation task. The models trained from synthetic data are difficult to generalize in some surveys. Recently, training 3D fault segmentation using sparse manual 2D slices is thought to yield promising results, but manual labeling has many false negative labels (abnormal annotations), which is detrimental to training and consequently to detection performance. Motivated to train 3D fault segmentation networks under sparse 2D labels while suppressing false negative labels, we analyze the training process gradient and propose the Mask Dice (MD) loss. Moreover, the fault is an edge feature, and current encoder-decoder architectures widely used for fault detection (e.g., U-shape network) are not conducive to edge representation. Consequently, Fault-Net is proposed, which is designed for the characteristics of faults, employs high-resolution propagation features, and embeds MultiScale Compression Fusion block to fuse multi-scale information, which allows the edge information to be fully preserved during propagation and fusion, thus enabling advanced performance via few computational resources. Experimental demonstrates that MD loss supports the inclusion of human experience in training and suppresses false negative labels therein, enabling baseline models to improve performance and generalize to more surveys. Fault-Net is capable to provide a more stable and reliable interpretation of faults, it uses extremely low computational resources and inference is significantly faster than other models. Our method indicates optimal performance in comparison with several mainstream methods.