In conventional remote sensing change detection (RS CD) procedures, extensive manual labeling for bi-temporal images is first required to maintain the performance of subsequent fully supervised training. However, pixel-level labeling for CD tasks is very complex and time-consuming. In this paper, we explore a novel self-supervised contrastive framework applicable to the RS CD task, which promotes the model to accurately capture spatial, structural, and semantic information through domain adapter and hierarchical contrastive head. The proposed SSLChange framework accomplishes self-learning only by taking a single-temporal sample and can be flexibly transferred to mainstream CD baselines. With self-supervised contrastive learning, feature representation pre-training can be performed directly based on the original data even without labeling. After a certain amount of labels are subsequently obtained, the pre-trained features will be aligned with the labels for fully supervised finetuning. Without introducing any additional data or labels, the performance of downstream baselines will experience a significant enhancement. Experimental results on 2 entire datasets and 6 diluted datasets show that our proposed SSLChange improves the performance and stability of CD baseline in data-limited situations. The code of SSLChange will be released at https: //github.com/MarsZhaoYT/SSLChange

Motivated by recent advances in industrial data communications, MW-class PV parks are increasingly utilizing communication-based coordination among different power converters to achieve accurate voltage regulation and stability, particularly during grid events. However, communication delays present in data transfer networks reduce coordination quality and deteriorate the ability of the PV park to respond in a timely manner. This paper investigates the impact of such delays on the voltage recovery performance of grid-tied converters in a PV park, coordinated by a centralized secondary voltage controller (SecVC), under different grid and delay conditions. In addition, to enhance the coordination quality, a local strategy for the converter-level controller is proposed to reduce the control sensitivity to the communication delays. The theoretical results obtained from nonlinear model and Monte-Carlo analysis are validated using a real communication setup between two Hardware-in-the-Loop devices (HIL), interfaced through a communication network emulator using the Modbus-TCP protocol.

The high penetration of inverter-based resources (IBRs) in power systems has brought challenges to system frequency stability due to reduced inertia. The conventional method for frequency stability assessment uses the centerof-inertia (CoI) frequency as a single indicator to represent the frequency throughout the entire system, which overlooks the spatio-temporal distribution of frequencies among multiple nodes during transients. The frequency distribution is further complicated by the differences in frequency regulation and synchronisation of synchronous generators (SGs) and IBRs, either grid-forming (GFM) or grid-following (GFL). To address this challenge, this paper develops a whole-system frequency response matrix in the frequency domain which accommodates the frequency regulation and synchronisation dynamics of all apparatuses in a unified framework. Based on this matrix, two frequency-domain metrics, namely spatio-temporal inertia and damping, are introduced to illustrate the frequency distribution characteristics. The spatial distribution is evaluated by using the matrix entries and the temporal distribution is evaluated by using the frequency spectra. The proposed model and metrics are validated via time-domain simulations on a modified 4-machine 2-area test system. The contributions of SGs and IBRs (GFM and GFL) to frequency regulation are examined by comparative analysis using the proposed metrics.

The fast and accurate interfacing of digital control platforms with power hardware constitutes a bottleneck in power electronic systems. The guidelines for the effective implementation of this interface are proprietary and not readily available to the designer. This paper presents a systematic approach for developing a mixed-domain control platform for high-frequency power converter applications. The proposed design methodology comprises three crucial steps. In the first step, the platform requirements are identified considering a wide range of switching frequency variations (up to tens of megahertz) found in power electronic applications. In the second step, a system architecture is derived, and its key building blocks and associated features are established based on the application requirements. These building blocks include serial ADCs and DACs, parallel ADCs and DACs for high-speed operation, protection circuits, PWM channels for converter gating, and digital communication features. In the third step, hardware components are selected for realizing different blocks based on a detailed comparison of the salient features of commercially available parts. A prototype of the proposed control platform is fabricated, and the experimental results are presented for validation. The results reinforce the criteria laid down in the design methodology. Further, the proposed control platform is compared with commercially available rapid control prototyping products and is shown to be economically effective for the desired features in typical power converter applications.

One important approach to improve the performance of the convolutional accelerator is to reduce its memory access latency. When the DDR memory connected to the accelerator is fixed, the memory access latency of the convolutional accelerator can be decreased by reducing the number of memory accesses or memory row conflicts, and using DMA to optimize memory access time intervals and so on. The convolutional process involves accessing a large amount of data with different sizes and types, which are usually stored in different memory rows. Accessing such data always causes a significant number of memory row conflicts, resulting in high memory access latency of the convolutional accelerator. This paper firstly analyzes the composition of DDR access latency for the convolutional accelerator and finds that the number of memory row conflicts is mainly decided by the number of loop tiling columns and padding size. Then, an OptTc method is proposed to calculate the optimal number of loop tiling columns. FPGA experiments of proposed convolution accelerator show that using the optimal number of loop tiling columns can reduce the memory access latency by 24% with negligible increase in FPGA resources in the case of VGG-16. The proposed OptTc method can effectively assist in the design of convolutional accelerators.

A fault-tolerant quantum computer must decode and correct errors faster than they appear to prevent exponential slowdown due to error correction. The Union-Find (UF) decoder is promising with an average time complexity slightly higher than O(d3). We report a distributed version of the UF decoder that exploits parallel computing resources for further speedup. Using an FPGA-based implementation, we empirically show that this distributed UF decoder has a sublinear average time complexity with regard to d, given O(d3) parallel computing resources. The decoding time per measurement round decreases as d increases, the first time for a quantum error decoder. The implementation employs a scalable architecture called Helios that organizes parallel computing resources into a hybrid tree-grid structure. Using a Xilinx VCU129 FPGA, we successfully implement d up to 21 with an average decoding time of 11.5 ns per measurement round under 0.1% phenomenological noise, and 23.7 ns for d = 17 under equivalent circuit-level noise. This performance is significantly faster than any existing decoder implementation. Furthermore, we show that Helios can optimize for resource efficiency by decoding d = 51 on a Xilinx VCU129 FPGA with an average latency of 544ns per measurement round.

Spinal cord injuries (SCI) substantially affect sensory, motor, and autonomous functions below the level of injury, reducing the independence and quality of life for affected individuals. Specifically, people with SCI between C5 and C7 cervical levels encounter limitations in voluntary finger and wrist flexion, reducing grasp capability. Compensatory strategies like tenodesis grasp, whereby wrist extension passively closes the fingers, remain; this is effective for small and light objects but insufficient for heavier ones. Typically, wearable assistive exoskeletons are designed to actuate a person’s fingers, however, such devices are sensitive to anatomical variability, such as hand size and joint contractures. The Dorsal Grasper is a wearable device designed to address this challenge by leveraging voluntary wrist extension and providing human-robot collaborative grasping capabilities with underactuated supernumerary fingers on the back of the hand. In this study, we introduce kinematic assessment methods that we use to show how the Dorsal Grasper expands the graspable workspace and reduces trunk motion, especially in situations where the use of a wheelchair restricts the individual’s posture. Our functionally relevant experiments with multiple SCI participants demonstrate the Dorsal Grasper’s potential as a versatile assistive solution for enhancing grasping capability in individuals with distinct SCI profiles.

The domain of industrial automation faces challenges such as shortened product life cycles, a shortage of skilled labor, and increasing complexity. Addressing these issues necessitates innovative solutions, one of which is the Digital Twin — a virtual counterpart of a physical asset. Central to the quality of a Digital Twin is the data it harnesses. While current Digital Twins draw data primarily from their corresponding physical assets, future interconnected production environments promise an influx of additional data from external devices. However, it remains uncertain how existing Digital Twins incorporate and leverage such data. In this systematic literature review, drawing from a pool of 1107 unique publications, we deeply analyzed 141 works to shed light on the data utilization in industrial Digital Twins. We categorized these publications based on Digital Twin types and classified them against various criteria regarding different characteristics of data. Our findings reveal that the majority of Digital Twins predominantly rely on structured data sourced directly from their associated assets, often employing proprietary integration methods. Facing the trends toward agile and interconnected production ecosystems as well as an increasing amount of unstructured data, we assert that current Digital Twins are not equipped to meet forthcoming demands in the industrial domain. Consequently, we propose necessary adaptations to fully unleash the potential of Digital Twins and outline future research fields, including automated data integration and evaluation.

The consumption of specialty coffee has increased around the world. Specialty coffee is free from impurities and defects. Specialty coffee is produced in smaller quantities, and its production process is hard and expensive. Traditional coffee beans have defects that affect the flavor of the coffee. It is essential to select the type of coffee to guarantee the best cost and quality. When a human manually classifies the type of coffee, there may be interference due to the human condition. This process has the disadvantage of being subjective. Few studies used Machine Learning methods for predicting specialty coffee classification by analyzing images. This article proposes a framework for classifying specialty coffees by applying Tiny Machine Learning techniques. We developed a model that can help accurately analyze and classify the coffee process and inform the quality of coffee without human interference. We achieved 100% accuracy, and we believe our model can be used effectively in the coffee industry.

The shift from fossil fuel to electric based propulsion in the waterborne transport sector has been sped up by recent policies aiming to reduce the sector emissions. This trend creates highly electrified vessels, with needs for energy storage systems (ESS) to satisfy the power demand affordably and to increase the on-board grid reliability and efficiency. Initial industry efforts have been put in the study and integration of high energy density ESS solutions, mainly electrochemical batteries. However, other innovative ESS, with different capabilities, have not been yet fully addressed. It is the case of Fast Response Energy Storage Systems (FRESS), such as Supercapacitors, Flywheels, or Superconducting Magnetic Energy Storage (SMES) devices. The EU granted project, POwer StoragE IN D OceaN (POSEIDON) will undertake the necessary activities for the marinization of the three mentioned FRESS. This study presents the design process followed in the POSEIDON project for the definition of an SMES suitable for maritime operation. First, the boundary conditions imposed by the marine environment, and the potential on-board applications of the SMES will be established. Next, the technological options: superconducting material, cooling system, coil fabrication and magnet topology which have been selected for this specific system will be presented.

This case study explores the practical approach to supply chain risk management as outlined in the McKinsey report (2019). It focuses on identifying, managing, and mitigating both known and unknown risks within the supply chain. The study details the implementation of a structured risk management process, including risk identification, development of a risk management framework, utilization of digital tools for real-time monitoring, and establishment of a cross-functional risk board for governance and review.Key findings highlight the effectiveness of this approach in mitigating risks, enhancing operational efficiency, and achieving cost savings. The use of digital tools provided real-time insights and early warnings, improving visibility and control over supply chain risks. The governance framework and regular reviews contributed to building resilience against potential disruptions.This case study provides valuable insights for industry practitioners and policymakers on the benefits of a structured approach to supply chain risk management, emphasizing the importance of strategic integration, continuous improvement, and operational synergy.

Unintended islanding refers to the disconnection of a grid section in a sequence of faults and uncontrolled continued operation of this section detached from the interconnected grid by chance. This condition should be avoided at the low-voltage level in particular, especially to ensure personal protection during repair work and to prevent damage to devices and equipment or blockages of the AWE. Therefore, current inverters are equipped with methods that recognize this condition and disconnect from the mains. It is believed that grid-forming technologies increase the risk of these islands developing and operating uncontrollably. There is therefore a need to investigate the effects of their future integration into the grid and, if necessary, to develop measures to safely detect and prevent the unwanted formation of islanded grids. The integration of grid-forming converters into the interconnected grid raises the question of how this will affect the ability of existing systems to detect islanding. This was examined in more detail in this study.

Data science is revolutionizing academia and industry, and there is a strong demand for a workforce fluent in data science. The availability of such courses has increased substantially in recent years. However, there are few rigorous curricula built on mathematical logic as the foundation of data science. The LogicDS Project aims to engage high school students from rural communities in an online data science course that integrates mathematics, statistics, and programming concepts into a single unified framework based on logic and reasoning. We developed a one-week data science course consisting of six lessons (a sampling of the full course offering) and recruited 110 participants. We collected pre- and post-intervention data and students' LMS activity log data to analyze their engagement. Results indicate that our Logic-Based framework for data science education effectively engages students from a variety of backgrounds; it was shown that they perceived the course as valuable for learning data science skills/concepts. Notably, our focus on the entropy analysis of student activity logs correlated to our other mixed methods analyses. This study provided insights for engaging K-12 students learning data science.

Polarimetric Synthetic Aperture Radar (PolSAR) images encompass valuable information that can facilitate extensive land cover interpretation and generate diverse output products. Extracting meaningful features from PolSAR data poses challenges distinct from those encountered in optical imagery. Deep Learning (DL) methods offer effective solutions for overcoming these challenges in PolSAR feature extraction. Convolutional Neural Networks (CNNs) play a crucial role in capturing PolSAR image characteristics by leveraging kernel capabilities to consider local information and the complexvalued nature of PolSAR data. In this study, a novel threebranch fusion of complex-valued CNN named Shallow to Deep Feature Fusion Network (SDF2Net) is proposed for PolSAR image classification. To validate the performance of the proposed method, classification results are compared against multiple stateof-the-art approaches using the Airborne Synthetic Aperture Radar (AIRSAR) datasets of Flevoland, San Francisco, and ESAR Oberpfaffenhofen dataset. The results indicate that the proposed approach demonstrates improvements in OA, with 1.3% and 0.8% enhancement for the AIRSAR datasets and 0.5% improvement for the ESAR dataset. The analyses conducted on the Flevoland data underscore the effectiveness of the SDF2Net model, revealing a promising OA of 96.01% even with only 1% sampling ratio. The source code is available at: https://github.com/mqalkhatib/SDF2Net

The Quantum Key Distribution (QKD) method leverages the principles of quantum mechanics to securely create and distribute private keys using quantum systems and an authenticated public classical channel. Despite offering informationtheoretical security, its physical implementations often suffer from unintended information leakage, known as side channels, which eavesdroppers can exploit to obtain the private key. This study introduces a classical-side-channel attack on the privacyamplification step of a general QKD protocol based on matrix hashing, utilizing machine-learning techniques to analyze powerconsumption leakage. Through multiple simulated scenarios, the study found that the gradient-boosting machine consistently outperformed other models, recovering the entire private key for high measuring-instrument sampling rates, regardless of the hashing matrix size and noise level tested. Additionally, the study proposes a strategy based on analyzing the confusion matrices of the model to facilitate a brute-force search for the critical feasible in the case of a non-perfect model. Furthermore, the study discusses countermeasures such as noise insertion, masking, and randomization techniques, potentially restoring the QKD protocol's information-theoretical security. This work demonstrates that machine-learning techniques can effectively characterize the leakages in a QKD implementation and create potent attacks.

A document by Bohan Chen. Click on the document to view its contents.

This paper discusses a method to optimize the Transformer model using the Arctic Puffin Optimization Algorithm (APO) for predicting and classifying users' emotional states. The experiment is divided into a main experiment and a comparison experiment. In the main experiment, the APO-optimized Transformer shows significant performance improvement, with the fitness curve stabilizing after two iterations. The model achieved 98.79% accuracy on the training set and 98.10% on the test set, with the accuracy and loss curves further validating its effectiveness. In the comparison experiment, we compared the proposed method with three basic machine learning models, with the random forest model performing best at 94.7% accuracy. However, the APO-optimized Transformer surpassed the random forest, achieving a 3.4% improvement with 98.10% accuracy. These results demonstrate the potential of the APO algorithm in deep learning and its effectiveness for social media sentiment analysis.

The advent of 6G wireless networks has the potential to unlock diverse applications of scalable autonomy. By advantageously coupling the individual and aggregated attributes of diverse multi-UAV fleets, a range of high-value applications such as logistics, enhanced disaster response, urban navigation, and surveillance can be significantly improved. However, enabling effective communication for knowledge fusion necessitates the intrinsic optimization of performance metrics like energy consumption, resource allocation, latency, and computational overheads to enhance autonomous efficiency. Furthermore, designing robust security features is essential to safeguarding privacy, control, and operational integrity. This paper explores a novel collaborative knowledge-sharing (KS) framework that leverages 6G and edge-computing capabilities to facilitate the cooperative training of decentralized machine learning models among multiple UAVs, without the need to transmit raw data. This framework aims to enhance the learning experience and operational efficiency of autonomous vehicles. The DECKS (distributed edge-based collaborative knowledge-sharing) architecture enables Federated Learning (FL) within UAV networks, allowing local models to be trained and shared among neighboring UAVs for creating global models. This promotes intelligent knowledge aggregation without a central entity, enhancing collaborative capabilities among autonomous vehicles. The DECKS architecture efficiently extracts and distributes collaborative shared experience to ground vehicles through edge and direct inference, reducing energy consumption, latency, and computational overhead. Our simulation analysis demonstrates that the DECKS architecture has the potential to reduce energy consumption by 70% in sensorless vehicles and improve autonomous vehicle learning performance by 15% compared to centralized approaches in a distributed environment. This improvement is achieved by comparing the efficiency of systems with and without aggregated knowledge.