Expensive Hardware Research Articles

Age of Processing-Based Data Offloading for Autonomous Vehicles in MultiRATs Open RAN

Today, vehicles use smart sensors to collect data from the road environment. This data is often processed onboard of the vehicles, using expensive hardware. Such onboard processing increases the vehicle’s cost, quickly drains its battery, and exhausts its computing resources. Therefore, offloading tasks onto the cloud is required. Still, data offloading is challenging due to low latency requirements for safe and reliable vehicle driving decisions. Moreover, age of processing was not considered in prior research dealing with low-latency offloading for autonomous vehicles. This paper proposes an age of processing-based offloading approach for autonomous vehicles using unsupervised machine learning, Multi-Radio Access Technologies (multi-RATs), and Edge Computing in Open Radio Access Network (O-RAN). We design a collaboration space of edge clouds to process data in proximity to autonomous vehicles. To reduce the variation in offloading delay, we propose a new communication planning approach that enables the vehicle to optimally preselect the available RATs such as Wi-Fi, LTE, or 5G to offload tasks to edge clouds when its local resources are insufficient. We formulate an optimization problem for age-based offloading that minimizes elapsed time from generating tasks and receiving computation output. To handle this non-convex problem, we develop a surrogate problem. Then, we use the Lagrangian method to transform the surrogate problem to unconstrained optimization problem and apply the dual decomposition method. The simulation results show that our approach significantly minimizes the age of processing in data offloading with 90.34% improvement over similar method.

EduTalk: An IoT Environment for Learning Computer Programming and Physics

This article proposes EduTalk, an out-of-the-box Internet of Things (IoT)-based smart learning environment for programming education. In particular, EduTalk enables the students to write VPython programs that render 3-D animations in the browser without installing extra software or using any specific (and typically expensive) hardware. EduTalk takes the user’s smartphone as a controller for cyber–physical interaction, which nicely integrates with learning of other core courses such as physics and mathematics. EduTalk allows building science exhibition projects by writing VPython programs to show 3-D animation, where the cost for EduTalk’s cyber–physical interaction is low and is almost maintenance free. The major contribution of this article is the IoT-based EduTalk proposal that subtly utilizes an IoT platform IoTtalk to conveniently generate cyber–physical interaction for learning how to program as well as learning core courses such as physics. The programming exercises can be easily extended to science exhibition projects and then the development of digital twin applications. A mechanism is provided to easily integrate GlowScript animation demos with EduTalk, which significantly simplifies the effort for teachers to prepare the lecturers. Finally, we show how data collected from EduTalk can be analyzed to improve learning design for cyber–physical interactive animation.

Performance of machine learning algorithms for detecting pine wilt disease infection using visible color imagery by UAV remote sensing

Automatic detection of pine wilt disease will be useful in forest management, as it can prevent the spread of epidemics and resulting huge economic losses. Previous studies have reported the precise pine wilt disease detection by combining machine learning algorithms and hyperspectral sensors with platform of unmanned aerial vehicles (UAVs). However, these approaches require expensive hardware and advanced processing techniques, rendering them unsuitable for practical use in on-site forest management. This study deals with the simplification of the automatic detection method, which enables practical countermeasures to be taken by forest management entities. An optimal method for detecting PWD using images obtained by a simple unmanned aerial vehicle (UAV)-mounted visible color sensor is presented in this study. The method combines six major machine learning algorithms: logistic regression, linear support vector machine, support vector machine, k-nearest neighbors, random forest, and artificial neural network, with both RGB and HSV color space datasets. The performance of the six algorithms was validated using the metrics of accuracy, precision, recall, F1-scores, and area under the receiver operating characteristic curve with their 95% confidence intervals. The experimental results show that the combination of the HSV color space dataset with the artificial neural network algorithm is the best performer algorithm (accuracy: 0.995). Moreover, logistic regression, linear support vector machine, support vector machine, and artificial neural network have equally high detection accuracy in the RGB color space (accuracy: 0.990, 0.990, 0.990, 0.985, respectively), with a low possibility of overfitting. These results indicate that the generally used visible color sensors mounted on UAV, which are available at a low cost, can be used for highly accurate detection.

Two-dimensional direction-of-arrival estimation based on time-domain-coding digital metasurface

Direction-of-arrival (DOA) estimation is one of the most critical technologies of radar, remote sensing, and wireless communications. The traditional DOA estimation is closely related to phased array antennas, which require complicated and expensive hardware and high-power consumption. Metasurface can manipulate electromagnetic waves without using massive transceivers, which makes it possible to realize antenna arrays in a cost-effective way. Here, we propose a strategy of DOA estimations by using a time-domain-coding digital metasurface and a single receiver. Specifically, the incident wave on the metasurface is modulated by the time-domain orthogonal codes impressed on the meta-atoms, and their amplitude and phase distributions are precisely retrieved from the signals detected by the receiver antenna. The effectiveness and accuracy of the proposed strategy are verified by experiments on a two-dimensional metasurface with individually addressable meta-atoms. The strategy features low cost and high flexibility and will facilitate various wireless applications.

S.M.A.R.T. F.U.S: Surrogate Model of Attenuation and Refraction in Transcranial Focused Ultrasound

Low-intensity focused ultrasound (LIFU) is an increasingly applied method for achieving non-invasive brain stimulation. However, transmission of ultrasound through the human skull can substantially affect focal point characteristics of LIFU, including dramatic attenuation in intensity and refraction of focal point location. These effects depend on a high-dimensional parameter space, making these effects difficult to estimate from previous work. Instead, focal point properties of LIFU experiments are often estimated using numerical simulation of LIFU sonication through skull. However, this procedure presents many entry barriers to even computationally savvy investigators and often requires expensive computational hardware, impeding LIFU research. We present a novel MATLAB toolbox (data: doi:10.5068/D1QD60; Matlab Scripts: https://doi.org/10.5281/zenodo.5811122) for rapidly estimating beam properties of LIFU transmitted through bone. Users provide specific values for frequency of LIFU, bone thickness, angle at which LIFU is applied, depth of the LIFU focal point, and diameter of the transducer used and receive an estimation of the degree of refraction/attenuation expected for the given parameters.

Library on-shelf book segmentation and recognition based on deep visual features

On-shelf book segmentation and recognition are crucial steps in library inventory management and daily operation. In this paper, a detailed investigation of related work is conducted. RFID and barcode-based solutions suffer from expensive hardware facilities and long-term maintenance. Digital Image processing and OCR techniques are flawed due to a lack of accuracy and robustness. On this basis, we propose a visual and non-character system utilizing deep learning methods to accomplish on-shelf book segmentation and recognition tasks. Firstly, book spine masks are extracted from the image of on-shelf books by instance segmentation model, followed by affine transformation to rectangle images. Secondly, a spine feature encoder is trained to learn the deep visual features of spine images. Finally, the book inventory search space is constructed and the similarity metric between spine visual representations is calculated to recognize the target book identity. To train the models we collect high-resolution datasets of 10k-level and develop a data annotation software accordingly. For validation, we design simulated scenarios of recognizing 3.6k IDs from 5.6k book spines and achieve a best top1 accuracy of 99.18% and top5 accuracy of 99.91%. Furthermore, we develop a prototype of a mobile library management robot with embedded edge intelligence. It can automatically perform on-shelf book image capturing, spine segmentation and recognition, and target book grasping workflow.

A Simple Elimination of the Thermal Convection Effect in NMR Diffusiometry Experiments.

Thermal convection is always present when the temperature of an NMR experiment is different from the ambient one. Most often, it falsifies the value of the diffusion coefficient determined by NMR diffusiometry using a PGSE NMR experiment. In spite of common belief, it acts not only at higher temperatures but also at temperatures lower than in the laboratory. Sodium alkyl-sulfate monomers and micelles in D2O solvent were used as model molecules measured at T = 319 K in order to show that thermal convection sometimes remains hidden in experiments. In this paper, we demonstrate that the increase in apparent diffusion coefficient with increasing diffusion time is a definite indicator of thermal convection. Extrapolation to zero diffusion time can also be used to obtain the real diffusion coefficient, likewise applying the less sensitive pulse sequences designed for flow compensation or the expensive hardware, e.g., sapphire or Shigemi NMR tubes, to decrease the temperature gradient. Further, we show experiments illustrating the effect of a long diffusion time in which the periodic changes of the echo intensity with gradient strength appear as predicted by theories.

A Remote Triggered Compound Parabolic Collector for Thermal Engineering Studies

Solar thermal energy systems are one of the most cost-effective renewable energy systems in use today. Engineering students study the design of these systems with the goal of learning how to design similar systems and perform research on improving the heating efficiency and overall operations. This paper elaborates on the design, construction, testing, and validation of a solar thermal system as a remote, open instrumentation lab, using two Compound Parabolic Collector (CPC) evacuated tube collectors with separate heating media. The lab allows for comparing heat transfer rates and collector efficiencies simultaneously for two fluids that have different thermal capacities. The heat patterns could be viewed using thermal cameras to analyze the CPC design. The unique feature of the system is its facility to control the lab remotely, as the setup is interfaced with instrumentation on a web server, thereby allowing students from geographically distant areas to access and perform experiments on the CPCs. A cumbersome lab with expensive hardware and outdoor requirements is thus made easy to perform and learn from via remote access. This remote methodology and hardware and IT architectures are especially pertinent and relevant in the blended and remote learning scenarios made common by the pandemic.

A thin metallo-dielectric stacked metamaterial as “add-on” for magnetic field enhancement in clinical MRI

Magnetic resonance imaging (MRI) is a widely used clinical diagnostic tool, which is based on the principle of nuclear magnetic resonance of hydrogen atoms in human body. The Larmor frequency of precession of the hydrogen atoms is determined by the strength of static magnetic field ( B0) of MRI. A higher B0 can directly improve signal-to-noise ratio (SNR) of MRI. However, this method involves expensive hardware installation, which could have adverse effects of tissue-heating and make MRI unsafe for patients with medical implants. Hence, efforts have been made to increase the SNR of MRI without increasing B0. An effective solution in this direction would be to boost the radiofrequency (RF) magnetic fields emitted by the body part undergoing scan, particularly by using metamaterials. The higher the received RF signal strength, the greater the SNR of MRI. For a metamaterial to be used as an “add-on” in commercial scanners, its dimensions need to be designed appropriately so that it fits in the available gap between the transceiver coil and the human body. In this article, a 10-mm-thick metallo-dielectric metamaterial is designed by a stacking of alternate square-shaped capacitive patches and inductive apertures for enhancing the RF magnetic flux density and hence, the SNR of a 1.5 T MRI system. The inter-layer electromagnetic coupling in the stacked structure is deployed for spatial localization of magnetic fields at the resonant frequency (∼64 MHz) which is equal to the Larmor frequency of 1.5 T MRI. An equivalent circuit model, comprising a lumped-element third order bandpass filter, validated the transmissivity characteristics of the metamaterial obtained using full-wave simulations. Magnetic flux density enhancement by a factor of 55 is obtained when the metamaterial add-on is placed between a surface coil and a bio-model of human head.

A deep neural networks-based image reconstruction algorithm for a reduced sensor model in large-scale tomography system

Image reconstruction for soft-field tomography is a highly nonlinear and ill-posed inverse problem. Owing to the highly complicated nature of soft-field, the reconstructed images are always poor in quality. One of the factors that affect image quality is the number of sensors in a tomography system. It is commonly assumed that increasing the number of sensors in a tomography system will improve the ill-posed condition in image reconstruction and hence improve image quality. However, as the number of sensors increases, challenges such as more complicated and expensive hardware, slower data acquisition rates, longer image reconstruction times, and larger sensitivity matrices will arise, resulting in a greater ill-posed condition. Since deep learning (DL) is capable of expressing complex nonlinear functions, the majority of research efforts have been directed toward developing a robust DL-based inverse solver for image reconstruction. However, no study has been conducted to solve the inverse problem and improve the quality of the reconstructed image using a reduced sensor model for a large-scale tomography system. This paper proposed an image reconstruction algorithm based on Deep Neural Networks (DNN) to investigate its feasibility in solving the ill-posed inverse problem caused by the reduced sensor model for a large-scale tomography system. The proposed DNN model is based on a supervised, feed-forward, fully connected, backpropagation network. It comprises an input layer, three hidden layers and an output layer. Also, it was trained using large data samples obtained from COMSOL simulation. The relationship between the scattered electromagnetic field measurement and the corresponding true electromagnetic field distribution vector is determined. During the image reconstruction process, the untrained scattered electromagnetic field measurement samples are used as inputs to the trained DNN model, and the model output is an estimate of the electromagnetic field distribution. The results show that the proposed DNN can accurately describe the distribution of electromagnetic field and boundary shape of phantom compared to traditional algorithms (LBP, FBP, Noser and Tikhonov), regardless of the size and number of phantoms within the monitoring area. Hence, the proposed DNN is more robust and has a high degree of generalization.

Guard Your Heart Silently

In recent years, particular attention has been devoted to continuous electrocardiogram (ECG) waveform monitoring due to its numerous applications. However, existing solutions require users to be confined to particular locations, rely on dedicated and expensive hardware, or require active user participation. The constrained recording conditions prevent them from being deployed in many practical application scenarios. In view of this, we present VibCardiogram, a continuous and reliable design for estimating ECG waveform shape via ubiquitous wrist-worn wearables; it renders a personal ECG waveform shape estimating system with prolonged recording time accessible to a larger sector of the population. Instead of adding auxiliary sensors to wearables, VibCardiogram leverages the widely integrated motion sensor to characterize cardiac activities, and interpret them to generate an alternative signal that has the same waveform shape as the ECG signal. Specifically, VibCardiogram extracts the cardiogenic body vibrations from noisy sensor data. As the waveform variability and inconstant period hinder the segmentation of cardiac cycles, VibCardiogram extracts features and realizes accurate segmentation via machine learning. To parse cardiac activities from vibration signals, we build a deep-learning pipeline associating the encoder-decoder framework and Generative Adversarial Networks. With dedicated construction and training, it can estimate the ECG waveform accurately. Experiments with 20 participants show that VibCardiogram can achieve an average estimation error of 5.989% for waveform amplitude estimation, which is within the 10% margins regulated by the American National Standards Institute. Moreover, the promising results further confirm that VibCardiogram effectively extracts Heart Rate Variability features and supports downstream applications.

Efficient evolutionary optimization using predictive auto-scaling in containerized environment

Solving complex real-world optimization problems is a computationally demanding task. To solve it efficiently and effectively, one must possess expert knowledge in various fields (problem domain knowledge, optimization, parallel and distributed computing) and appropriate expensive software and hardware resources. In this regard, we present a cloud-native, container-based distributed optimization framework that enables efficient and cost-effective optimization over platforms such as Amazon ECS/EKS, Azure AKS, and on-premise Kubernetes. The solution consists of dozens of microservices scaled out using a specially developed PETAS Auto-scaler based on predictive analytics. Existing schedulers, whether Kubernetes or commercial, do not take into account the specifics of optimization based on evolutionary algorithms. Therefore, their performance is not optimal in terms of results’ delivery time and cloud infrastructure costs. The proposed PETAS Auto-scaler elastically maintains an adequate number of worker pods following the exact pace dictated by the demands of the optimization process. We evaluate the proposed framework’s performance using two real-world computationally demanding optimizations. The first use case belongs to the manufacturing domain and involves optimization of the transportation pallets for train parts. The second use case belongs to the field of automated machine learning and includes neural architecture search and hyperparameter optimization. The results indicate an IaaS cost savings of up to 49% can be achieved, with almost unchanged result delivery time.

Mind Games: A Web-Based Multiplayer Brain-Computer Interface Game

The advancement of consumer-grade Brain-Computer Interface (BCI) devices has led to the use of brain wave data in disciplines beyond neuroscience. While BCI devices are studied holistically in some domains, the potential remains unexplored in other applicable domains, such as gaming. The majority of BCI games are single-player, only work on a single platform, or require expensive hardware that is difficult to use. To address these challenges, we propose Mind Games: a competitive multiplayer tug-of-war game played entirely with brain wave data in a web browser with a low-cost consumer-grade BCI device. Unlike other BCI games, Mind Games is platform-independent, has a quick setup time, and uses the latest web-based technologies: web-Bluetooth and JavaScript frameworks for networking and signal processing. This new cross-platform BCI pipeline enables physiological computing applications to be more accessible, leading the future of online multiplayer BCI games.

Adaptive Dead Time Compensation for Continuous Cross-Period Single Phase Shift Control of Dual Active Bridge Converters

The load transient response of a power converter is an important property, especially in a test equipment application. The dynamic behavior of a dual active bridge converter mainly depends on the switching frequency, the transformer leakage inductance, and the capacitor bank size. Using the Continuous Cross-Period Single Phase-Shift control, an excellent load transient response can be achieved without the need for expensive hardware components. However, in practical applications, the switch delay and the required inserted dead time distort the current waveform, increasing the electromagnetic noise. i

Barriers to SME Computerization in Developing Countries: Evidence from SMEs in North Kivu, Democratic Republic of Congo

SMEs are widely recognized as the development enablers in developing countries. They contribute to job creation and, in so doing, to poverty reduction. To remain competitive, SMEs use several tools, including computerized information systems. However, given their characteristics, several constraints hinder their computerization; these constraints are even more accentuated in developing countries. This study attempted to identify the barriers to computerization as outlined in the literature and to ascertain them in the DRC context. In order to achieve this, we examined the relevant literature and carried out a survey of 53 respondents, principally local ICT stakeholders. Using descriptive statistics, we established that Congolese SMEs face the same computerization challenges identified in the literature. These constraints include the local SME characteristics - poor financial resources, low internal and external expertise, and low management involvement -, high infrastructure costs - expensive software and hardware - and the availability of the technological infrastructure - poor internet access -. The Congolese digital ecosystem stakeholders can, each at their own level, use these results to assist the digitalization of Congolese SMEs and thus make them more agile in the achievement of their goals. Keywords: ICT, Barriers, Computerization, DRC, SME.

Discriminating different materials by means of vibrations

Material characterization and discrimination is of interest for multiple applications, ranging from mechanical engineering to medical and industrial sectors. Despite the need for automated systems, the majority of the existing approaches necessitate expensive and bulky hardware that cannot be used outside ad-hoc laboratories. In this work, we propose a novel technique for discriminating between different materials and detecting intra-material variations using active stimulation through vibration and machine learning techniques. A voice-coil actuator and a tri-axial accelerometer are used for generating and sampling mechanical vibration propagated through the materials. Results of the present analysis confirm the effectiveness of the proposed approach. Processing a mechanical vibration signal that propagates through a material by means of a neural network is a viable means for material classification. This holds not only for distinguishing materials having gross differences, but also for detecting whether a material underwent some slight changes in its structure. In addition, mechanical vibrations at 500 Hz demonstrated an ability to provide a compact and meaningful representation of the data, sufficient to categorize 8 different materials, and to distinguish reference materials from other defective materials, with an average accuracy greater than 90%.

Novel streak-resolving algorithm for particle streak velocimetry

In this paper, a novel streak-resolving algorithm is proposed for particle streak velocimetry (PSV) to resolve velocity fields for a single image frame with multiple particle streaks. A model streak function, which was based on the temporal integral of the particle image intensity moving along its trajectory during the exposure period, was approximated using a multivariable least-square fit procedure to reconstruct the streak information and the corresponding particle trajectory. Lagrangian tracking was achieved, and the velocities were evaluated by differentiating the resolved particle trajectory with respect to time. Two types of flows, accelerating parallel flow and the rotational flow of Hill’s vortex, were used to generate synthetic streaks for the performance tests. Three types of relative error were defined and used to evaluate the performance of the algorithm in terms of statistical mean and standard deviation (SD) errors. The accuracy of the fitted streak parameters, such as particle image intensity and diameter, were also evaluated and compared. The results reveal that the error and SD were low if the image noise is below 1.0%; for noise levels of 5.0%, the error was up to 10% with an SD of up to 12%. The error and SD of the particle image intensity and particle image diameter for both flow types were also 0%–7% for clean and up to 12% for noisy images. The processing results for experimental streak images of flow past a cylinder reveal that these images can be resolved using the proposed algorithm with a residual mean error of 4.38 and an SD of 9.48. These results suggest that the proposed novel approach can be used to resolve velocity fields with only a single image frame and without expensive hardware for high-speed imaging and thus is suitable for diverse applications.

Vision-based mapping of lane semantics and topology for intelligent vehicles

High-definition map is an essential tool for route measurement, planning and navigation of intelligent vehicles. Yet its creation is still a persisting challenge, especially in creating the semantic and topology layer of the map based on visual sensing. However, current semantic mapping approaches do not consider the map applicability in navigation tasks while the topology mapping approaches face the issues of limited location accuracy or expensive hardware cost. In this paper, we propose a joint mapping framework for both semantic and topology layers, which are learned in a lane-level and based on a monocular camera sensor and an on-board GPS positioning device. A map management approach “RoadSegDict” is also proposed to support the efficient updating of semantic map in a crowdsourced manner. Moreover, a new dataset is proposed, which includes a variety of lane structures with detailed semantic and topology annotations.

Enabling Cost-Effective, SLO-Aware Machine Learning Inference Serving on Public Cloud

The remarkable advances of Machine Learning (ML) have spurred an increasing demand for ML- as-a-Service on public cloud: developers train and publish ML models as online services to provide low-latency inference for dynamic queries. The primary challenge of ML model serving is to meet the response-time Service-Level Objectives (SLOs) of inference workloads while minimizing serving cost. In this article, we proposes MArk (Model Ark), a general-purpose inference serving system, to tackle the dual challenge of SLO compliance and cost effectiveness. MArk employs three design choices tailored to inference workload. First, MArk dynamically batches requests and opportunistically serves them using expensive hardware accelerators (e.g., GPU) for improved performance-cost ratio. Second, instead of relying on feedback control scaling or over-provisioning to serve dynamic workload, which can be too slow or too expensive, MArk employs predictive autoscaling to hide the provisioning latency at low cost. Third, given the stateless nature of inference serving, MArk exploits the flexible, yet costly serverless instances to cover occasional load spikes that are hard to predict. We evaluated the performance of MArk using several state-of-the-art ML models trained in TensorFlow, MXNet, and Keras. Compared with the premier industrial ML serving platform SageMaker, MArk reduces the serving cost up to <inline-formula><tex-math notation="LaTeX">$7.8\times$</tex-math></inline-formula> while achieving even better latency performance.

Comparison of Bone Segmentation Software over Different Anatomical Parts

Three-dimensional bone shape reconstruction is a fundamental step for any subject-specific musculo-skeletal model. Typically, medical images are processed to reconstruct bone surfaces via slice-by-slice contour identification. Freeware software packages are available, but commercial ones must be used for the necessary certification in clinics. The commercial software packages also imply expensive hardware and demanding training, but offer valuable tools. The aim of the present work is to report the performance of five commercial software packages (Mimics®, AmiraTM, D2PTM, SimplewareTM, and Segment 3D PrintTM), particularly the time to import and to create the model, the number of triangles of the mesh, and the STL file size. DICOM files of three different computed tomography scans from five different human anatomical areas were utilized for bone shape reconstruction by using each of these packages. The same operator and the same hosting hardware were used for these analyses. The computational time was found to be different between the packages analyzed, probably because of the pre-processing implied in this operation. The longer “time-to-import” observed in one software is likely due to the volume rendering during uploading. A similar number of triangles per megabyte (approximately 20 thousand) was observed for the five commercial packages. The present work showed the good performance of these software packages, with the main features being better than those analyzed previously in freeware packages.