Search

Eye movements reflect the visual process of humans’ perception and cognition. In the field of medical imaging, the diagnosis rendered by radiologists is closely related to their eye movements when reading radiological images. It is beneficial to study the eye movements of radiologists to improve the diagnostic performance. However, existing studies are mainly focused on the radiologists’ fixations but rarely on their saccade patterns. Moreover, these studies are almost based on limited datasets. In this paper, we present a quantitative study of the gaze behavior of radiologists from the perspective of saccade patterns on a large-scale dataset. The dataset comprises of the eye-tracking data of 10 expert radiologists reading 196 mammograms. By analyzing the saccade amplitude, direction, and bias of radiologists, we found that radiologists have specific saccade patterns in image reading and the saccade patterns are significantly affected by the different reading phases, working experience, and orientations of the mammograms.

Monocular depth estimation is an essential technique for tasks like 3D reconstruction. Although many works have emerged in recent years, they can be improved by better utilizing the multi-scale information of the input images, which is proved to be one of the keys in generating high-quality depth estimations. In this paper, we propose a new monocular depth estimation method named HMA-Depth, in which we follow the encoder-decoder scheme and combine several techniques such as skip connections and the atrous spatial pyramid pooling. To obtain more precise local information from the image while keeping a good understanding of the global context, a hierarchical multi-scale attention module is adopted and its outputs are combined to generate the final output that is with both good details and good overall accuracy. Experimental results on two commonly-used datasets prove that HMA-Depth can outperform the existing approaches. Code is available 1 1 https://github.com/saranew/HMADepth., MVA2021,17th International Conference on Machine Vision Applications,July 25–27, 2021, Fully Online

An essential prerequisite for unleashing the potential of supervised deep learning algorithms in the area of 3D scene understanding is the availability of large-scale and richly annotated datasets. However, publicly available datasets are either in relative small spatial scales or have limited semantic annotations due to the expensive cost of data acquisition and data annotation, which severely limits the development of fine-grained semantic understanding in the context of 3D point clouds. In this paper, we present an urban-scale photogrammetric point cloud dataset with nearly three billion richly annotated points, which is three times the number of labeled points than the existing largest photogrammetric point cloud dataset. Our dataset consists of large areas from three UK cities, covering about 7.6 km^2 of the city landscape. In the dataset, each 3D point is labeled as one of 13 semantic classes. We extensively evaluate the performance of state-of-the-art algorithms on our dataset and provide a comprehensive analysis of the results. In particular, we identify several key challenges towards urban-scale point cloud understanding. The dataset is available at https://github.com/QingyongHu/SensatUrban., Comment: CVPR 2021, Code: https://github.com/QingyongHu/SensatUrban

It is meaningful to recommend exercises to students in an online education system with a large amount of learning resource. Many recommendation methods usually rely on strategy in the recommendation system in order to predict an exercise score. In this paper, we compare different information entropy index in the exercise recommendation. These index consider how well exercise matches the student’s knowledge ability. In order to compare different methods, we introduce an interactive platform for dynamic exercise recommendation. We conduct a set of exercise recommendation experiments, compare the effect of different index and find the optimal index in the experiment.

Since the inception of industrialization, power system has been an indispensable aspect of economy. With the progression of time, technology has impalpably commingled into our lifestyle. Alongside blooming technologies, energy demand is proliferating and power companies are begetting energy at their best to quench it. Growing reliance on power system has brought its quality into more advertence. Various electronic devices and topologies have been invented to enhance power quality and reliability; numerous others are still underway. During the course, power system has grown to an intricate network of sources, loads and control devices, leading to various issues such as transmission congestion and high losses. This paper discusses ways to ameliorate congestion and gives an overview of relationship between our present energy resources and ecological threats like global warming. Moreover, it points out various power system problems such as energy losses and transients. The necessity of FACTS devices has also been elaborated alongside their classification and comparison. Finally, numerous topologies and optimization methods proposed in the technical literature have been classified and analyzed to alleviate power system conundrums, and a glimpse into future energy trends is presented.

Photo response non-uniformity (PRNU) noise is a sensor pattern noise characterizing the imaging device. It has been broadly used in the literature for image authentication and source camera identification. The abundant information that the PRNU carries in terms of the frequency content makes it unique, and therefore suitable for identifying the source camera and detecting forgeries in digital images. However, PRNU estimation from smartphone videos is a challenging process due to the presence of frame-dependent information (very dark/very textured), as well as other non-unique noise components and distortions due to lossy compression. In this paper, we propose an approach that considers only the non-textured frames in estimating the PRNU because its estimation in highly textured images has been proven to be inaccurate in image forensics. Furthermore, lossy compression distortions tend to affect mainly the textured and high activity regions and consequently weakens the presence of the PRNU in such areas. The proposed technique uses a number of texture measures obtained from the Grey Level Cooccurrence Matrix (GLCM) prior to an unsupervised learning process that splits the feature space through training video frames into two different sub-spaces, i.e., the textured space and the non-textured space. Non-textured video frames are filtered out and used for estimating the PRNU. Experimental results on a public video dataset captured by various smartphone devices have shown a significant gain obtained with the proposed approach over the conventional state-of-the-art approach.

Electric vehicles (EVs) play a key role in transport electrification and decarbonizing the society. EVs are becoming popular due to the advancement of drivetrain and battery technologies, with the support from plummeting costs. However, many countries are facing with challenges to accommodate large-scale adoption of electric vehicles due to the great amount of electricity required from the grid for charging. Current research focuses on developing incentives and tariff structure to encourage EV drivers to charge and discharge at appropriate times. However, relying on driver's behavior can be a risky decision for network operators. A regional high electricity demand at a short instance can cause severe technical challenges for the distribution network, including thermal and voltage limit violations. This paper presents and compares the EV development for the UK and China. A research agenda is proposed to consider how large-scale energy storage would benefit the distribution network for rapid charging of electric vehicles.

In this paper, we present a Markov chain based model that can be used for easy estimation of the Vehicle to Vehicle (V2V) message generation rates in a highway environment based on just the traffic flow rate and the average vehicle speed on the highway. This allows for a faster estimation of the overall channel load than using a simulation environment to do the same. Our model considers the effects of Decentralized Congestion Control based on Transmit Rate Control (DCC-TRC) on Cooperative Awareness Message (CAM) generations. The model is evaluated by comparing the results obtained with the message generation rates achieved for a highway traffic scenario in a simulation environment based on Artery and SUMO. Comparing the results shows that the Cumulative Distributive Function (CDF) of the message generation rates estimated by our model is fairly accurate as they fall within the 95% confidence interval of the CDF obtained from the simulation.

In this paper, a new type of low-cost free-standing all-metal frequency selective surface (AM-FSS) is studied and presented. The AM-FSS is developed using an extremely thin sheet of metal having a total thickness of λ 0 /80, where λ 0 is the free-space wavelength at the operating frequency. The fundamental element of the AM-FSS is a square ring of area 0.45λ 0 × 0.45λ 0 and having fork-type stubs extending from four sides towards the center of the ring. The transmission and reflection characteristics of the AM-FSS are predicted using a commercial electromagnetic simulator. Higher-order bandpass FSSs are realized using two and three layers of the proposed AM-FSSs. The independent polarization characteristic of the AM-FSS is demonstrated by predicting the performance of the three-layered AM-FSS for two orthogonal TE and TM polarizations. The proposed AM-FSS does not use any commercial laminates, making it extremely attractive for low-cost and high-power microwave applications.

In this paper, we propose a novel frequency selective surface (FSS) antenna for monitoring the moisture content (MC) in X-band (8–12 GHz). In the proposed sensor, the sample under test acts as the substrate of the antenna sandwiched between an array of square split ring resonators (SSRR) at the top plate and a ground plate at the bottom. The center frequency is chosen as 10 GHz. The width, length, and thickness of the FSS are 21 mm, 50 mm, and 0.017 mm, respectively. The compactness, sensitivity, and simplicity of the proposed antenna make it a suitable candidate as a portable microwave sensor for moisture content monitoring., 2021 IEEE 19th International Symposium on Antenna Technology and Applied Electromagnetics (ANTEM), August 8-11, 2021, Winnipeg, MB, Canada

This paper presents a novel multi-modal learning approach for automated skill characterization of obstetric ultrasound operators using heterogeneous spatio-temporal sensory cues, namely, scan video, eye-tracking data, and pupillometric data, acquired in the clinical environment. We address pertinent challenges such as combining heterogeneous, small-scale and variable-length sequential datasets, to learn deep convolutional neural networks in real-world scenarios. We propose spatial encoding for multi-modal analysis using sonography standard plane images, spatial gaze maps, gaze trajectory images, and pupillary response images. We present and compare five multi-modal learning network architectures using late, intermediate, hybrid, and tensor fusion. We build models for the Heart and the Brain scanning tasks, and performance evaluation suggests that multi-modal learning networks outperform uni-modal networks, with the best-performing model achieving accuracies of 82.4% (Brain task) and 76.4% (Heart task) for the operator skill classification problem.

Fetal brain magnetic resonance imaging (MRI) offers exquisite images of the developing brain but is not suitable for second-trimester anomaly screening, for which ultrasound (US) is employed. Although expert sonographers are adept at reading US images, MR images which closely resemble anatomical images are much easier for non-experts to interpret. Thus in this paper we propose to generate MR-like images directly from clinical US images. In medical image analysis such a capability is potentially useful as well, for instance for automatic US-MRI registration and fusion. The proposed model is end-to-end trainable and self-supervised without any external annotations. Specifically, based on an assumption that the US and MRI data share a similar anatomical latent space, we first utilise a network to extract the shared latent features, which are then used for MRI synthesis. Since paired data is unavailable for our study (and rare in practice), pixel-level constraints are infeasible to apply. We instead propose to enforce the distributions to be statistically indistinguishable, by adversarial learning in both the image domain and feature space. To regularise the anatomical structures between US and MRI during synthesis, we further propose an adversarial structural constraint. A new cross-modal attention technique is proposed to utilise non-local spatial information, by encouraging multi-modal knowledge fusion and propagation. We extend the approach to consider the case where 3D auxiliary information (e.g., 3D neighbours and a 3D location index) from volumetric data is also available, and show that this improves image synthesis. The proposed approach is evaluated quantitatively and qualitatively with comparison to real fetal MR images and other approaches to synthesis, demonstrating its feasibility of synthesising realistic MR images., IEEE Transactions on Medical Imaging 2020

DC system protection is more challenging than that for AC system due to the rapid rate of rise of the fault current and absence of natural current zero-crossing in DC systems. Superconducting fault current limiter (SFCL) in DC systems is a promising technology to reduce the fault current level and the rate of rise of the fault current, and also SFCLs have no resistance during normal operation. In this paper, the behaviors of an SFCL coil are investigated under both low impedance and high impedance fault conditions in DC systems. In the low impedance fault condition system, the SFCL coil performs effective limitation of the fault current level under different prospective fault current levels. The application of SFCLs with limited inductance in the DC system can be a potential solution to effectively suppress the fault current under low impedance short-circuit faults. The SFCL coil under the high impedance fault condition can only limit the prospective fault current when it is much higher than the critical current of the coil.

In this paper, an umbrella-shaped patch antenna for future millimeter-wave applications is presented. The proposed antenna resonates at multiple frequency bands at 28 GHz, 38 GHz, and 55 GHz (V-band) that have been globally allocated for 5G communications systems. The proposed antenna is designed using Rogers RT/duroid 5870, with a thickness of 0.79 mm. The antenna has an overall size of 8 mm × 8 mm with simple geometrical structure, compact size, multiband operation, high gain, and efficiency.

Employing unmanned aerial vehicles (UAVs) in millimeter-wave (mmWave) networks as relays has emerged as an appealing solution to assist remote or blocked communication nodes. In this case, the network security becomes a great challenge due to the presence of malicious eavesdroppers. In this paper, we perform a secrecy analysis for a UAV-based mmWave relaying network. We first investigate the relaying scheme without jamming where the UAV decodes and forwards the information from the source to the destination with malicious eavesdropping. Furthermore, to enhance the secrecy performance, we propose a cooperative jamming scheme via utilizing the destination and an external UAV to cooperatively disrupt the eavesdroppers at the two stages of relaying, respectively. Using the probability of line-of-sight (LoS) between the UAV and ground nodes, the three-dimensional (3D) antenna gain, and the Nakagami-m small-scale fading model, the secrecy outage probability (SOP) of the two schemes with and without jamming is analyzed. Closed-form expressions for the SOP of the two schemes are obtained by employing the Gauss-Chebyshev quadrature. Simulation results are presented to validate the theoretical expressions of SOP and to show the effectiveness of the proposed schemes.

Accurate recognition of non-driving activity (NDA) is important for the design of intelligent Human Machine Interface to achieve a smooth and safe control transition in the conditionally automated driving vehicle. However, some characteristics of such activities like limited-extent movement and similar background pose a challenge to the existing 3D convolutional neural network (CNN) based action recognition methods. In this paper, we propose a dual-stream 3D residual network, named D3D ResNet, to enhance the learning of spatio-temporal representation and improve the activity recognition performance. Specifically, a parallel 2-stream structure is introduced to focus on the learning of short-time spatial representation and small-region temporal representation. A 2-feed driver behaviour monitoring framework is further build to classify 4 types of NDAs and 2 types of driving behaviour based on the drivers head and hand movement. A novel NDA dataset has been constructed for the evaluation, where the proposed D3D ResNet achieves 83.35% average accuracy, at least 5% above three selected state-of-the-art methods. Furthermore, this study investigates the spatio-temporal features learned in the hidden layer through the saliency map, which explains the superiority of the proposed model on the selected NDAs.

Deep Learning (DL) has been widely proposed for botnet attack detection in Internet of Things (IoT) networks. However, the traditional Centralized DL (CDL) method cannot be used to detect previously unknown (zero-day) botnet attack without breaching the data privacy rights of the users. In this paper, we propose Federated Deep Learning (FDL) method for zero-day botnet attack detection to avoid data privacy leakage in IoT edge devices. In this method, an optimal Deep Neural Network (DNN) architecture is employed for network traffic classification. A model parameter server remotely coordinates the independent training of the DNN models in multiple IoT edge devices, while Federated Averaging (FedAvg) algorithm is used to aggregate local model updates. A global DNN model is produced after a number of communication rounds between the model parameter server and the IoT edge devices. Zero-day botnet attack scenarios in IoT edge devices is simulated with the Bot-IoT and N-BaIoT data sets. Experiment results show that FDL model: (a) detects zero-day botnet attacks with high classification performance; (b) guarantees data privacy and security; (c) has low communication overhead (d) requires low memory space for the storage of training data; and (e) has low network latency. Therefore, FDL method outperformed CDL, Localized DL, and Distributed DL methods in this application scenario.

Wireless Internet of Things (IoT) devices will be deployed to enable applications such as sensing and actuation. These devices are typically resource-constrained and are unable to perform resource-intensive computations. Therefore, these jobs need to be offloaded to resource-rich nodes at the edge of the IoT network for execution. However, the timeliness and correctness of edge nodes may not be trusted (such as during high network load or attack). In this paper, we look at the applicability of trust for successful offloading. Traditionally, trust is computed at the application level, with suitable mechanisms to adjust for factors such as recency. However, these do not work well in IoT networks due to resource constraints. We propose a novel device called Trust Tracker (denoted by Σ) that provides higher-level applications with up-to-date trust information of the resource-rich nodes. We prove impossibility results regarding computation offloading and show that Σ is necessary and sufficient for correct offloading. We show that, Σ cannot be implemented even in a synchronous network and we compute the probability of offloading to a bad node, which we show to be negligible when a majority of nodes are correct. We perform a small-scale deployment to demonstrate our approach.\ud

The Second Generation Digital Video Broadcasting Terrestrial (DVB-T2) standard and its light version (DVB-T2-Lite) offer very flexible system configurations including multiple-input single-output (MISO) transmission technology to improve the performance of the Single Frequency Networks (SFNs). In this paper, a measurement-based study of the DVB-T2/T2-Lite MISO performance under different mobile and portable transmission scenarios is presented. The outputs of the study, evaluated by both objective and subjective ways, show that the influence of different mobile and portable transmission conditions on the performance of DVB-T2/T2-Lite MISO broadcasting is not negligible.

Providing early diagnosis of cerebral palsy (CP) is key to enhancing the developmental outcomes for those affected. Diagnostic tools such as the General Movements Assessment (GMA), have produced promising results in early diagnosis, however these manual methods can be laborious. In this paper, we propose a new framework for the automated classification of infant body movements, based upon the GMA, which unlike previous methods, also incorporates a visualization framework to aid with interpretability. Our proposed framework segments extracted features to detect the presence of Fidgety Movements (FMs) associated with the GMA spatiotemporally. These features are then used to identify the body-parts with the greatest contribution towards a classification decision and highlight the related body-part segment providing visual feedback to the user. We quantitatively compare the proposed framework's classification performance with several other methods from the literature and qualitatively evaluate the visualization's veracity. Our experimental results show that the proposed method performs more robustly than comparable techniques in this setting whilst simultaneously providing relevant visual interpretability., Comment: Proceedings of the 2021 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI), accepted, 2021

Due to the proliferation of edge computing, cloud providers have started offering compute nodes at the edge of the network in addition to traditional compute nodes in data centers. So far, various systems have been proposed for processing Internet of Things (IoT) data on both edge and cloud compute nodes in order to reduce the communication latency. However, such systems do not typically consider that the network bandwidth between an edge node and a cloud node can be orders of magnitude higher than the bandwidth between an IoT device and a cloud node. As a result, the IoT data are commonly sent selectively to either edge or cloud nodes disregarding alternative network paths through edge nodes, which may have higher network bandwidth, and lower communication latency. To avoid this, in this paper we analyze the latency of sending data to edge and cloud compute nodes of cloud providers. Based on this analysis, we propose edgeRouting which routes the data through the closest edge compute node. By doing that, edgeRouting exploits both the low propagation delay of nodes at the edge, and the high bandwidth among edge and cloud compute nodes of cloud providers. To evaluate our approach, we perform experiments on a real-world setup with nearby and remote compute nodes of a cloud provider, and we show that edgeRouting reduces the communication latency by up to 55% compared to alternative methods.

The rapid development of new semiconductor materials and devices has highlighted the need for compact modeling and circuit simulation tools that can be easily adapted to accommodate emerging technologies. In most instances device modeling tools employ non-linear behavioural sources and Verilog-A modules for model prototype construction. This paper is concerned with the properties and application of modular user-defined/plugin library toolkit that combines the best features of behavioural source and Verilog-A modeling practice while encouraging user extensions. The toolkit has been implemented as a Qucs/Qucs-S/Xyce modular library that is loadable on demand. To demonstrate its capabilities and flexibility a series of compact device models are introduced and their simulated performance presented and evaluated

Deep Learning models, applied to a sensor-based Human Activity Recognition task, usually require vast amounts of annotated time-series data to extract robust features. However, annotating signals coming from wearable sensors can be a tedious and, often, not so intuitive process, that requires specialized tools and predefined scenarios, making it an expensive and time-consuming task. This paper combines one of the most recent advances in Self-Supervised Leaning (SSL), namely a SimCLR framework, with a powerful transformer-based encoder to introduce a Contrastive Self-supervised learning approach to Sensor-based Human Activity Recognition (CSSHAR) that learns feature representations from unlabeled sensory data. Extensive experiments conducted on three widely used public datasets have shown that the proposed method outperforms recent SSL models. Moreover, CSSHAR is capable of extracting more robust features than the identical supervised transformer when transferring knowledge from one dataset to another as well as when very limited amounts of annotated data are available.

The development of Unmanned Aerial Systems (UASs) continuously improves and advances the technology, which is a key enabler for many new applications, such as autonomous Beyond Visual Line of Sight (BVLOS) operations. However, ensuring a sufficient level of safety and performance of an UAS can be a challenging task, since it requires systematic experimental validation in order verify and document the reliability, robustness, and fault tolerance of the UAS and its critical components. In this paper we propose the UAV Auto Test Framework (UAVAT Framework), a framework for easy, systematic, and efficient experimental validation of the reliability, fault tolerance, and robustness of a multirotor small Unmanned Aerial Systems (sUAS) and its software and hardware components. We describe the hardware and software used for the UAVAT Framework setup, which consists of a Motion Capture (MoCap) system, a multirotor sUAS, and a tethered power system. In addition, we introduce the concept of test modules, a 'plug-and-play' software component with a set of recommended guidelines for defining testing configurations and enabling easy reuse and distribution of software components for sUAS testing. The capabilities of the UAVAT Framework are demonstrated by presenting the results from two developed test modules targeting endurance testing of a multirotor sUAS and Fault Detection (FD) for abnormal behaviour detection.

Safe autonomous Beyond Visual Line of Sight (BVLOS) flight is critical for the widespread exploitation of Unmanned Aerial Systems (UASs). Satisfying safety legislation requires documenting compliance with a raft of substantial requirements, a step which very few commercially available UASs have yet taken. Assessments under BVLOS conditions are required - but illegal for unassessed UASs. We propose to enable safe BVLOS flight tests using an Independent Failsafe System (IFS), which can be added to an Unmanned Aerial Vehicle (UAV) and enable a pilot to remotely monitor flights and terminate them safely via parachute if necessary. Our IFS has been designed for use under the European Union (EU) UAS legislation. In this paper, we describe this IFS and define a systematic approach for integrating its use into the Specific Operations Risk Assessment (SORA) process necessary for BVLOS certification. The use of this IFS to safely fly an (otherwise) unsafe UAV has been approved by the Danish Civilian Aviation Authority (CAA) for BVLOS flights with a range of small multirotor UAVs. The results of experimental evaluations in closed airspace are presented.

The amount of data in real-time, such as time series and streaming data, available today continues to grow. Being able to analyze this data the moment it arrives can bring an immense added value. However, it also requires a lot of computational effort and new acceleration techniques. As a possible solution to this problem, this paper proposes a hardware architecture for Typicality and Eccentricity Data Analytic (TEDA) algorithm implemented on Field Programmable Gate Arrays (FPGA) for use in data streaming anomaly detection. TEDA is based on a new approach to outlier detection in the data stream context. The suggested design has a full parallel input of N elements and a 3-stage pipelined architecture to reduce the critical path and thus optimize the throughput. In order to validate the proposals, results of the occupation and throughput of the proposed hardware are presented. The design reached a speed of up to 693x, compared to other software platforms, with a throughput of up to 10.96 MSPs (Mega Sample Per second), using a small portion of the target FPGA resources. Besides, the bit accurate simulation results are also presented. This work is a pioneer in the hardware implementation of the TEDA technique in FPGA. The project aims to Xilinx Virtex-6 xc6vlx240t-1ff1156 as the target FPGA.

As a fundamental task in natural language processing, sentence semantic matching (SSM) is critical yet challenging due to difficulties in learning expressive sentence representation while capturing complex interactions between sentences. Recent work has shown the great potential of deep neural models in improving the performance of SSM task. However, existing work usually employs recurrent neural networks (RNNs) or 1D (one-dimensional) convolutional neural networks (CNNs) to learn sentence representation, leading to limited performance improvement. Benefiting from the multi-dimensional structure, 2D convolutional neural networks are expected to be more powerful to learn expressive sentence representation by capturing the implicit inter-sentence interactions and thus can further improve the performance of SSM. To this end, in this paper, we propose a novel sentence semantic matching model named Hierarchical CNN based on Dimension-augmented Representation (HiDR). In HiDR, first, bidirectional long short-term memory networks (LSTMs) are utilized to generate dimension-augmented representation for each of the input sentences; then, a hierarchical 2D CNN is devised to learn sentence representation while capturing the inter-sentence interactions, followed by a prediction layer based on sigmoid function to output the matching degree between sentences. To evaluate the performance of our proposed model, we conducted extensive experiments on two public real-world data sets. The empirical results show that HiDR has achieved remarkable results, which demonstrates either better or comparable performance w.r.t. BERT-based models.

Zero-shot learning (ZSL) aims to recognize instances belonging to unseen categories which are not available at training time. Previous ZSL models learn a projection function from the visual feature space to a semantic space which contains a description of the categories. The semantic attributes are often correlated with each other at the semantic space and it is not appropriate to learn them independently. Existing ZSL methods are designed to work on complete descriptions of the semantic attributes. However, because these attributes are human-designed values, they might be incomplete or contains noisy values which may affect the recognition performance of many existing ZSL models. This paper proposes a novel zero-shot learning approach (ZSL-MSA) to handle missing and noisy semantic attributes during the training process. Significantly, the proposed method learns a supplementary attribute matrix by exploiting the attribute correlation. The proposed method also learns the relevant feature coefficients in the projection matrix to identify the correlated attribute space. Th proposed method also adopts l 1 regularization norm to select the relevant sparse features. A constrained optimization function is formulated and solved using the accelerated proximal gradient method. Extensive experiments on three benchmark datasets using ZSL and generalized ZSL demonstrate the effectiveness of the proposed method.

Autonomous vehicles operating in dynamic environments are required to account for other traffic participants. By interpreting sensor information and assessing the collision risk with vehicles, cyclists, and pedestrians, near-misses and accidents can be prevented. Moreover, by explaining risk factors to developers and engineers the overall safety of autonomous driving can be increased in future deployments. In this paper, we have designed, developed, and evaluated an approach for predicting the collision risk with other road users based on a planar 2D collision model. To this end, we have trained interpretable machine learning models to classify and predict the risk of collisions on a range of features extracted from sensor data. Further, we present methods for inferring and explaining the factors mostly contributing to the risk. Using counterfactual inference, our approach allows us to determine the factors which highly influence the risk and should in turn be minimised. Experimental results on real-world driving data show that collision risk can be effectively predicted and explained for different time horizons as well as different types of traffic participants such as cars, cyclists and pedestrians.

Autonomous systems increasingly rely on on-board computation to avoid the latency overheads of offloading to more powerful remote computing. This requires the integration of hardware accelerators to handle the complex computations demanded by date-intensive sensors. FPGAs offer hardware acceleration with ample flexibility and interfacing capabilities when paired with general purpose processors, with the ability to reconfigure at runtime using partial reconfiguration. Managing dynamic hardware is complex and has been left to designers to address in an ad-hoc manner without first-class integration in autonomous software frameworks. This paper presents an abstracted runtime for managing adaptation of FPGA accelerators, including partial reconfiguration and parametric changes, that presents as a typical interface used in autonomous software systems. We present a demonstration using the Robot Operating System (ROS), showing negligible latency overhead as a result of the abstraction.

Voice over IP (VoIP) is finding its way into several applications, but its security concerns still remain. This paper\ud shows how a new self-adaptive steganographic method can ensure the security of covert VoIP communications over the\ud Internet. In this study an Active Voice Period Detection algorithm is devised for PCM codec to detect whether a VoIP packet\ud carries active or inactive voice data, and the data embedding location in a VoIP stream is chosen randomly according to random\ud sequences generated from a logistic chaotic map. The initial parameters of the chaotic map and the selection of where to\ud embed the message are negotiated between the communicating parties. Steganography experiments on active and inactive\ud voice periods were carried out using a VoIP communications system. Performance evaluation and security analysis indicates\ud that the proposed VoIP steganographic scheme can withstand statistical detection, and achieve secure real-time covert\ud communications with high speech quality and negligible signal distortion.

Light-Fidelity (LiFi) access points (APs) can be deployed anywhere to support the exponential growth of various lightweight hand-held and smart internet-of-things (IoT) devices and applications. IoT traffic is often sporadic and consists of heterogeneous bursty flows, which may affect the performance of supported wireless services. This paper develops an optimal data rate-based slicing scheme that dynamically virtualizes a networked LiFi AP downlink channel bandwidth according to the data rate demands of users and the requested services. We formulate a game optimization problem that captures the relation between three main entities: a LiFi attocellular network infrastructure provider (InP), mobile virtual network operators (MVNOs), and IoT or wireless users. A software-defined networking (SDN) controller enables the InP to reconfigure the LiFi AP downlink channel slices to MVNOs, based on their service-level agreements (SLAs) and the traffic profile of IoT services. Simulation results show that the proposed AP resources slicing scheme is promising for IoT applications.

In this paper, a decode-and-forward (DF) short-packet relaying model is developed to achieve timely status updates for intelligent monitoring within the Internet of Things (IoT), where the status updates generated at an IoT device are delivered to a remote server with the aid of a relay in both half-duplex (HD) and full-duplex (FD) modes. To characterise the data freshness of status updates, we exploit the age of information (AoI) as a metric, which is defined as the time elapsed since the generation of the latest successfully decoded status update. The average AoI is formulated and minimised for both HD-DF and FD-DF relaying IoT networks in finite blocklength regime. For the HD-DF relaying, we introduce a perfect approximation of the average AoI to solve the problem of average AoI minimisation with the optimal blocklengths in two phases. For the FD-DF relaying, we propose an iterative algorithm to solve the problem of average AoI minimisation by optimising the relay’s transmit power and the blocklength. Illustrative numerical results not only substantiate the validity of our proposed algorithms, but also provide useful references for the IoT monitoring network design, specifically for the transmit power thresholds at the IoT device and the relay.

Industry 4.0 heavily builds on massive deployment of Industrial Internet of Things (IIoT) devices to monitor every aspect of the manufacturing processes. Since the data gathered by these devices impact the output of critical processes, identity management and communications security are critical aspects, which commonly rely on the deployment of X.509 certificates. Nevertheless, the provisioning and management of individual certificates for a high number of IIoT devices involves important challenges. In this paper, we present a solution to improve the management of digital certificates in IIoT environments, which relies on partially delegating the certificate enrolment process to an edge server. However, in order to preserve end-to-end security, private keys are never delegated. Additionally, for the protection of the communications between the edge server and the IIoT devices, an approach based on Identity Based Cryptography is deployed. The proposed solution considers also the issuance of very short-lived certificates, which reduces the risk of using expired or compromised certificates, and avoids the necessity of implementing performance expensive protocols such as Online Certificate Status Protocol (OCSP). The proposed solution has been successfully tested as an efficient identity management solution for IIoT environments in a real industrial environment. This work was supported in part by the Spanish Ministry of Science and Innovation through the National Towards zeRo toUch nEtwork and services for beyond 5G (TRUE-5G) Project under Grant PID2019-108713RB-C53, in part by the European Commission through the Electronic Components and Systems for European Leadership-Joint Undertaking (ECSEL-JU) 2018 Program under the framework of key enabling technologies for safe and autonomous drones' applications (COMP4DRONES) Project under Grant 826610, with the national financing from France, Spain, Italy, The Netherlands, Austria, Czech, Belgium, and Latvia, in part by the Ayudas Cervera para Centros Tecnologicos Grant of the Spanish Centre for the Development of Industrial Technology (CDTI) through the Project EGIDA under Grant CER-20191012, and in part by the Basque Country Government through the Creating Trust in the Industrial Digital Transformation (TRUSTIND) ELKARTEK Program Project under Grant KK-2020/00054.

Modern manufacturing relies heavily on digital technologies, and the recent changes in the manufacturing environment are the reflection of the advancements in information and communication technologies. Web-based Digital Twin (WDT) will constitute the future of manufacturing giving a greater potential of process/product data interaction, where Digital Twin functions on a web browser and connects to its Physical Twin to exchange data. To this end, the research work on WDT is still in the first stages. Therefore, the current paper presents a framework for developing WDT taking into account the possibility of utilising WDT for education, research and industrial applications. A case study adopted from a mini-scale assembly line is used to illustrate the proposed concept.\ud

Accurate and efficient electromagnetic transient (EMT) simulation of various types of solid-state transformers (SST) is extremely time-consuming due to the complex module structure, flexible topology connections, large number of electrical nodes and simulation time-steps limited in the range of micro-seconds. Therefore, it is urgent to develop the EMT equivalent modelling and fast simulation of SSTs for system level studies. Taking the modular multilevel converter (MMC) based SST as an example, this paper proposes an accelerated EMT model which focuses on the equivalence of the dual active bridge (DAB) based high-frequency link (HFL) in the SST. Compared with the existing algorithms, two critical factors of the proposed method that contribute the most to the efficiency improvement are the preprocessing of the nodal admittance equation and the conversion of the short-circuit admittance parameters. The proposed model is verified in PSCAD/EMTDC by comparing it with the detailed EMT model. The results show that the accelerated model is one to two orders of magnitude faster than the detailed model without sacrificing the accuracy. The experiment validation also confirms the validity of the proposed model.

The operation of Unmanned Aerial Systems (UAS) is widely recognised to be limited globally by challenges associated with gaining regulatory approval for flight Beyond Visual Line of Sight (BVLOS) from the UAS Remote Pilot. This challenge extends from unmanned aircraft flights having to follow the same ‘see and avoid’ regulatory principles with respect to collision avoidance as for manned aircraft. Due to the technical challenges of UAS and Remote Pilots being adequately informed of potential traffic threats, this requirement effectively prohibits BVLOS UAS flight in uncontrolled airspace, unless a specific UAS operational airspace is segregated from manned aviation traffic, often achieved by use of a Temporary Danger Area (TDA) or other spatial arrangements. The UK Civilian Aviation Authority (CAA) has defined a Detect and Avoid (DAA) framework for operators of UAS to follow in order to demonstrate effective collision avoidance capability, and hence the ability to satisfy the ‘see and avoid’ requirement. The National BVLOS Experimentation Corridor (NBEC) is an initiative to create a drone experimentation facility that incorporates a range of surveillance and navigation information sources, including radars, data fusion, and operational procedures in order to demonstrate a capable DAA System. The NBEC is part located within an active Airodrome Traffic Zone (ATZ) at Cranfield Airport, which further creates the opportunity to develop and test systems and procedures together with an operational Air Traffic Control (ATC) unit. This allows for manned and unmanned traffic to be integrated from both systems and procedural perspectives inside segregated airspace in a first stage, and then subsequently transiting to/from non-segregated airspace. The NBEC provides the environment in which a number of challenges can be addressed. This paper discusses the lack of target performance parameters, the methodology for gaining regulatory approval for non-segregated BVLOS flights and for defining peformance parameters for counter UAS (cUAS).

Cyber-Physical Systems (CPSs) are getting increasingly complex and generate large amounts of data. Analyzing such data provides us an insight into a given system. The digital twin concept emerges as an attempt to seamlessly integrate the data and insight in order to improve system performances. It enables applications such as visualization, monitoring, state estimation, and self-adaptation. In this paper, we demonstrate the construction of a digital twin exemplified by an incubator system, including the benefits and challenges of each application. The result is a description of the building blocks of a digital twin.

State-of-the-art battery-free RFID systems attach multiple RFID tags to an object and exploit their RF phase to estimate its three-dimensional (3D) orientation. However, the measured RF phase may be inaccurate because each tag's signal fingerprint (i.e., RSSI and RF Phase) is distorted by multipath interference and electromagnetic interaction between neighboring tags. In this paper, we propose RF-Orien3D that minimizes these interferences for accurate 3D orientation recognition only using two RFID tags. The electromagnet interference modifies the radiation pattern and modulation factor of each tag in the two-element tag array, which can be estimated to compensate for the distortion in RFID fingerprints. To deal with the multipath impact, we simulate multipath noise to generate huge amounts of RFID fingerprints and use them to pre-train a convolutional neural network (CNN). Then we only collect dozens of actual samples to fine-tune the CNN for multipath-tolerant orientation recognition. The experiments show RF-Orien3D recognizes a two-tag labeled object's 2D orientation with the angular error of about 16° and its 3D orientation (azimuth and elevation) with the errors of about 29° and 11° in low/rich multipath scenarios.

Computation over multiple access channel (CoMAC) scheme provides a promising solution to future large-scale wireless networks by utilizing the superposition property of the wireless channel to compute a class of functions with a summation structure (e.g., mean, norm, etc.). However, its implementation usually requires all nodes’ channel state information (CSI) and its performance is limited by the channel condition of the worst node. In order to avoid massive CSI aggregation and improve the limited performance, we propose an automatic repeat request (ARQ)-aided CoMAC scheme in this paper. The transmitters and signaling procedures are designed to achieve the tradeoff between the achievable function rate and the transmission delay. The corresponding performance of the proposed ARQ-aided CoMAC scheme and the traditional ARQ-aided communication scheme are compared for both homogeneous networks and heterogeneous networks. By optimizing the ARQ level, we further maximize the achievable function rate of the proposed scheme. Asymptotic closed-form expressions are derived by resorting to the extreme value theory and point mass approximation. Monte Carlo simulations are given to illustrate and verify the performance of the proposed designs.

This paper demonstrates a step on the transition towards a digital twin for a desktop version of an agricultural robot. This includes implementation of motor control and in-door localisation capabilities for the robot. Data from the physical twin is streamed to a co-simulation based on the Functional-Mockup Interface, in which a digital twin simulates the robot’s movement. Safety constraints are established inside the digital twin and a proof of concept communication is enabled when such constraints are violated.

In preparation of the Call for Proposals on JPEG Pleno Holography, multiple exploration studies are ongoing to define the general procedure for performance assessment of proposed coding solutions. The performance of proposals will be compared against JPEG 2000 and H.265/HEVC intra coding, both serving as anchor codecs. In this paper, we report on the results of a dynamic subjective visual quality assessment procedure. A subset of 8 holograms, selected from the JPEG Pleno Database, is compressed both in hologram and object plane at three different bit rates. Chosen bit rates vary per hologram depending on their content characteristics. Then, for each hologram and its compressed versions, pseudo-video sequences are generated from the reconstructed views along a scan path that involves focus and viewing angle changes and that is hologram specific. A double stimulus simultaneous test, combined with a 5-level impairment-scale scoring protocol is deployed where videos created from the reference and the decoded holograms are visualized side by side on a professional 4K display. Results demonstrate that this test procedure requires deep scenes and sufficient scene complexity throughout the depth stack to allow for adequate stress testing of the codecs under test, particularly those solutions that adhere to compression in the object plane.

Cybersecurity education and training are essential prerequisites of achieving a secure and privacy-friendly digital environment. Both professionals and the general public widely acknowledge the need for high-quality university education programs and professional training courses. However, guides, recommendations, practical tools, and good examples that could help institutions design appropriate cybersecurity programs are still missing. In particular, a comprehensive method to identify skills needed by cybersecurity work roles offered on the job market is missing. This paper aims to provide practical tools and strategies to help higher education providers design good cybersecurity curricula. First, we analyze the content of 89 existing study programs worldwide, collect recommendations of renowned institutions within and outside the EU, and provide a comprehensive survey accompanied by a dynamic web application called Education Map. Based on the knowledge about the current state in cybersecurity education, we design the SPARTA Cybersecurity Skills Framework that provides the currently missing link between work roles and required expertise and shows how to develop a curriculum that reflects job market requirements. Finally, we provide a practical tool that implements the framework and helps education and training providers design new study programs and analyze existing ones by considering the requirements of cybersecurity work roles.

Emission of greenhouse gases and scarcity of fossil fuels have put the focus of the scientific community, industry and society on the electric vehicle (EV). In order to reduce CO2 emissions, cutting-edge policies and regulations are being imposed worldwide, where the use of EVs is being encouraged. In the best of scenarios reaching 245 million EVs by 2030 is expected. Extensive use of EV-s requires the installation of a wide grid of charging stations and it is very important to stablish the best charging power topology in terms of efficiency and impact in the grid. This paper presents a review of the most relevant issues in EV charging station power topologies. This review includes the impact of the battery technology, currently existing standards and proposals for power converters in the charging stations. In this review process, some disadvantages of current chargers have been identified, such as poor efficiency and power factor. To solve these limitations, five unidirectional three-phase rectifier topologies have been proposed for fast EV charging stations that enhance the current situation of chargers. Simulation results show that all the proposed topologies improve the power factor issue without penalizing efficiency. The topologies with the best overall performance are the Vienna 6-switch and the Vienna T-type rectifier. These two converters achieve high efficiency and power factor, and they allow a better distribution of losses among semiconductors, which significantly increase the life-cycle of the semiconductor devices and the reliability of the converter. This work was supported in part by the Government of the Basque Country through the Fund for Research Groups of the Basque University System under Grant IT978-16, in part by the Research Program ELKARTEK under Project ENSOL2-KK-2020/00077 and Project HARVESTGEN-KK-2020/00113, in part by the Ministerio de Ciencia e Innovacion of Spain under Project PID2020-115126RB-I00, and in part by the FEDER Funds. Document

The digitalization has become a leading trend for developing a stable economy through industry and firm digital transformation. However, there is no unified recipe for the successful implementation of these technologies. This work aims to define Critical Success Factors of digitalization for the manufacturing industry and compare them with Kazakhstani conditions. The methodology is based on the literature review and survey of stakeholders. This paper presents the concept of critical success factors identification and defines the necessity of further research.

The Internet of Medical Things (IoMT) couples the rapid growth of Internet of things (IoT) technologies with smart health systems, leveraging wireless battery-operated devices for remote health monitoring. Since 2019, a surge in the number of COVID-19 patients has increased rapidly, leading to increased strain on hospital resources and leaving some urgent patients behind. This is substantial cause to transform interactive health treatment into intelligent healthcare using edge computing and artificial intelligence (AI) techniques. However, running sophisticated AI-based edge computing techniques on IoT devices with limited battery is not sustainable. Hence, addressing the trade-off between energy-efficiency and smart AI techniques is imperative to maximize the device's lifetime. This paper proposes a Multi-Modal Reinforcement Learning (MMRL) algorithm that will help maximize the IoT device's lifetime using adaptive data compression, energy-efficient communication, and minimum latency, particularly for emergency events. The results showed a 500% longer battery life than the state-of-the-art algorithms in addition to high adaptability to different conditions. This work was made possible by NPRP grant NPRP12S-0305-190231 from the Qatar National Research Fund (a member of Qatar Foundation).

Increasing integration of converter-interfaced renewable power generation challenges the stability of the frequency in the modern power systems due to the lack of inertia in this type of generators. In this regard, this paper establishes a control method for controlling multilevel converters as an interface between renewable power generation and the power grid, which enables them to operate as voltage-controlled sources, thus grid-forming capabilities like conventional synchronous generators (SGs). In fact, the proposed control technique mimics the behavior of synchronous generators in the control loop of the neutral point clamped (NPC) converter. Consequently, supportive functionalities for frequency stability i.e., inertia and frequency/power oscillation damping will be provided by the multilevel converter as a virtual synchronous generator (VSG). Simulation results in Matlab/Simulink verifies the well-designed inertia and damping features for the interfaced converter as well as appropriate performance of multilevel converter in reducing total harmonic distortion (THD).

Additive manufacturing (AM) has been revolutionizing the manufacturing industry due to its ability to significantly reduce waste and produce components with intricate shapes. Laser Ultrasonics (LU) is a non-contact and couplant free method to generate and detect ultrasound. LU can accommodate complex component shapes; thus, it has the potential to provide a reliable in-process inspection method for AM components. In recent years the development of Laser Induced Phased Arrays (LIPAs) helped overcome the inherently low signal amplitudes of LU at the non-destructive, thermoelastic regime. In this paper, the Full Matrix Capture data acquisition method is used and a LIPA of 68 elements is synthesized in post processing. The Total Focusing Method imaging algorithm is applied for ultrasonic imaging. The technique is demonstrated on a highly scattering titanium alloy Wire Arc Additive Manufactured (WAAM) component producing high quality ultrasonic images, accurately imaging defects at depths up to 10mm below the inspection surface.

The share of renewables in the power system is increasing rapidly. Large offshore wind power plants (OWPPs) are developed at a high pace and conventional fossil fuel-based plants are decommissioned. Consequently, there will be a risk of insufficient amount of power plants providing black start functionality for system restoration after a black out. This paper proposes a STATCOM with a battery energy storage that is located at the point of common connection to an OWPP that together can provide a reliable black start service to the power grid. The concept is demonstrated by using time domain simulations in PSCAD. The STATCOM functionality provides fast and dynamic reactive power management and the battery unit provides active power balancing capability to maintain the frequency within a tolerable range specified by the system operator. The simulation results fulfill the success criteria for the black start and confirm its feasibility for practical implementation.

Autorefraction consists of the automatic sensing of three parameters – spherical error, cylindrical error, slope of the principal meridian – that describe the deviation of the focusing properties of an ametropic eye with respect to an emmetropic state. Low-cost autorefractors would be highly desirable in resource-poor settings, for the stratification of patients between those who can be treated in the community and those who need to be referred to specialist care. In the present paper, we describe the implementation of an autorefractor based on projecting patterns onto the retina of an eye and observing the projected pattern through an ophthalmoscopic camera configuration coaxial with the projection path. Tunable optics in the coaxial path, combined with appropriate image processing, allows determination of the three parameters. The simplicity and performance of the setup, measured on an eye simulator, shows promise towards clinical use in the community. Further work is needed to confirm the performance in vivo.