Search

Accurate traffic prediction is vital for effective traffic management during hurricane evacuation. This article proposes a predictive modeling system that integrates multilayer perceptron and long short-term memory models to capture both long-term congestion patterns and short-term speed patterns. Leveraging various input variables, including archived traffic data, spatial-temporal road network information, and hurricane forecast data, the framework is designed to address challenges posed by heterogeneous human behaviors, limited evacuation data, and hurricane event uncertainties. Deployed in a real-world traffic prediction system in Louisiana, USA, the model achieved an 82% accuracy in predicting long-term congestion states over a 6-h period during a seven-day hurricane-impacted duration. The short-term speed prediction model exhibited mean absolute percentage errors ranging from 7% to 13% across evacuation horizons from 1 to 6 h. The evaluation results underscore the model’s potential to enhance traffic management during hurricane evacuations, and real-world deployment highlights its adaptability and scalability in diverse hurricane scenarios within extensive transportation networks. [ABSTRACT FROM AUTHOR]

This article explores the critical challenges and limitations of modern communication-based train control (CBTC) systems, particularly focusing on the dynamic and uncertain nature of train–ground communication. The concept of the Age of Information (AoI) is introduced, highlighting the discrepancies between the actual and derived states used for control, which can compromise system performance and safety. Existing solutions, such as wireless local area networking technologies and reinforcement learning handoff schemes, are discussed, emphasizing their contributions to reducing handoff latency and improving the freshness of information between the train and access points. However, their limitations in adequately addressing issues related to the AoI are critically examined. The article argues for the development of novel control approaches that consider these effects to minimize the information gap and enhance the overall performance of CBTC systems. [ABSTRACT FROM AUTHOR]

Generally, municipal water supply companies use manual collection and laboratory analysis for water quality testing. However, these methods have limitations such as lack of real-time information, inability to sample the entire water supply, and high costs. Therefore, continuous, real-time water quality monitoring is crucial for public health protection and ensuring that the whole water supply network is monitored. This paper proposes an Internet of Things (IoT) platform for the measurement of consumption and the quality of drinking water in rural or semi-rural environments. Data collected through temperature, flow, potential of hydrogen (pH), turbidity, and Oxidation Reduction Potential (ORP) sensors is exchanged with a database through a long-range wireless communication protocol. Two mobile applications and one desktop application were also developed, with the purpose of being used by simple users, technicians, and network administrators respectively. The presented implementation process includes the design of the hardware surrounding the ESP32 microcontroller and its mounted peripherals, as well as the software run by the microcontroller and the mobile devices. A prototype system was built and tested under controlled conditions, successfully recognizing an increase in water turbidity and its unsuitability when contaminated with different agents. This method may prove to be a financially advantageous solution for rural, semi-rural, and even urban environments when used with groups of data collection nodes, helping significantly in the upkeep and surveillance of the water supply network.Highlights: The IoT platform is equipped with sensors that measure water consumption as well as turbidity, temperature, and pH. Tested under controlled conditions, detecting different signs of water contamination. A low-cost option for continuous water monitoring, particularly in resource-limited environments. [ABSTRACT FROM AUTHOR]

Durational action timed automata (daTAs) are state transition systems like timed automata (TAs) that capture information regarding the concurrent execution of actions and their durations using maximality-based semantics. As the underlying semantics of daTAs are infinite due to the modeling of time progress, conventional model checking techniques become impractical for systems specified using daTAs. Therefore, a finite abstract representation of daTA behavior is required to enable model checking for such system specifications. For that, we propose a finite abstraction of the underlying semantics of a daTA-like region abstraction of timed automata. In addition, we highlight the unique benefits of daTAs by illustrating that they enable the verification of properties concerning concurrency and action duration that cannot be verified using the traditional TA model. We demonstrate mathematically that the number of states in durational action timed automata becomes significantly smaller than the number of states in timed automata as the number of actions increases, confirming the efficiency of daTAs in modeling behavior with action durations. [ABSTRACT FROM AUTHOR]

In this paper, a novel multimodal large language model-based fault detection and diagnosis framework that addresses the limitations of traditional fault detection and diagnosis approaches is proposed. The proposed framework leverages the Generative Pre-trained Transformer-4-Preview model to improve its scalability, generalizability, and efficiency in handling complex systems and various fault scenarios. Moreover, synthetic datasets generated via large language models augment the knowledge base and enhance the accuracy of fault detection and diagnosis of imbalanced scenarios. In the framework, a hybrid architecture that integrates online and offline processing, combining real-time data streams with fine-tuned large language models for dynamic, accurate, and context-aware fault detection suited to industrial settings, particularly focusing on security concerns, is introduced. This comprehensive approach aims to address traditional fault detection and diagnosis challenges and advance the field toward more adaptive and efficient fault diagnosis systems. This paper presents a detailed literature review, including a detailed taxonomy of fault detection and diagnosis methods and their applications across various industrial domains. This study discusses case study results and model comparisons, exploring the implications for future developments in industrial fault detection and diagnosis systems within Industry 4.0 technologies. [ABSTRACT FROM AUTHOR]

The paper describes low-cost hardware-based analog and digital real-time circuit simulators for the development of power electronics control circuits. During the process of designing and developing digital control circuits for power electronics systems, preliminary verification of control algorithms is required. For this purpose, software simulators such as Pspice, Psim, Matlab-Simulink, and many others are commonly used. Afterward, the developed control algorithm is implemented in the digital control system. For further verification of the implemented control algorithms, a hardware-based analog or digital simulator can be utilized. The paper presents the author's proposed analog simulators. In the digital version of the simulator, TMS320F28388D microcontroller with 200 MHz clock was used. These simulators have demonstrated their usefulness in the development of power electronics systems. [ABSTRACT FROM AUTHOR]

Copyright of Przegląd Elektrotechniczny is the property of Przeglad Elektrotechniczny and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Cache partitioning is a technique to reduce interference among tasks running on the processors with shared caches. To make this technique effective, cache segments should be allocated to tasks that will benefit the most from having their data and instructions stored in the cache. The requests for cached data and instructions can be retrieved faster from the cache memory instead of fetching them from the main memory, thereby reducing overall execution time. The existing partitioning schemes for real-time systems divide the available cache among the tasks to guarantee their schedulability as the sole and primary optimization criterion. However, it is also preferable, particularly in systems with power constraints or mixed criticalities where low- and high-criticality workloads are executing alongside, to reduce the total cache usage for real-time tasks. Cache minimization as part of design space exploration can also help in achieving optimal system performance and resource utilization in embedded systems. In this paper, we develop optimization algorithms for cache partitioning that, besides ensuring schedulability, also minimize cache usage. We consider both preemptive and non-preemptive scheduling policies on single-processor systems with fixed- and dynamic-priority scheduling algorithms (Rate Monotonic (RM) and Earliest Deadline First (EDF), respectively). For preemptive scheduling, we formulate the problem as an integer quadratically constrained program and propose an efficient heuristic achieving near-optimal solutions. For non-preemptive scheduling, we combine linear and binary search techniques with different fixed-priority schedulability tests and Quick Processor-demand Analysis (QPA) for EDF. Our experiments based on synthetic task sets with parameters from real-world embedded applications show that the proposed heuristic: (i) achieves an average optimality gap of 0.79% within 0.1× run time of a mathematical programming solver and (ii) reduces average cache usage by 39.15% compared to existing cache partitioning approaches. Besides, we find that for large task sets with high utilization, non-preemptive scheduling can use less cache than preemptive to guarantee schedulability. [ABSTRACT FROM AUTHOR]

A two‐step framework that integrates real‐time data collection with time series forecasting models for predicting traffic volume is proposed. In the first step, the framework utilizes live highway surveillance video data and YOLO‐v7 object detector to construct accurate traffic volume data. In the second step, an ARIMA–LSTM time series model is applied to forecast future traffic volumes. Experimental results show that YOLO‐v7 achieved a vehicle detection accuracy of over 93.30%, ensuring high precision in traffic volume data construction. The ARIMA–LSTM model demonstrated superior performance in traffic volume prediction, with a mean squared error of 87.97, root mean squared error of 10,388.57, and mean absolute error of 101.39. YOLO‐v7's detection speed of 7.8 ms per frame further validates the feasibility of real‐time data construction. The findings indicate that the combination of YOLO‐v7 for vehicle detection and ARIMA–LSTM for traffic prediction is highly effective, offering a significant reduction in training time compared to more complex deep learning models while maintaining high prediction accuracy. This research presents a unified solution for traffic data collection and prediction, enhancing transportation infrastructure planning and optimizing traffic flow. Future work will focus on extending the prediction intervals and further refining the models to improve performance. [ABSTRACT FROM AUTHOR]

The integration of digital twins (DTs) in healthcare is critical but remains limited in real‐time patient monitoring due to challenges in achieving low‐latency telemetry transmission and efficient resource management. This paper addresses these limitations by presenting a novel cloud‐based DT framework that optimises real‐time healthcare monitoring, providing a timely solution for critical healthcare needs. The framework incorporates a Pyomo‐based dynamic optimisation model, which reduces telemetry latency by 32% and improves response time by 52%, surpassing existing systems. Leveraging low‐cost, low‐latency multimodal sensors, the system continuously monitors critical physiological parameters, including SpO2, heart rate, and body temperature, enabling proactive health interventions. A DT definition language (Digital Twin Definition Language)‐based time series analysis and twin graph platform further enhance sensor connectivity and scalability. Additionally, the integration of machine learning (ML) strengthens predictive accuracy, achieving 98% real‐time accuracy and 99.58% under cross‐validation (cv = 20) using the XGBoost algorithm. Empirical results demonstrate substantial improvements in processing time, system stability, and learning capacity, with real‐time predictions completed in 17 ms. This framework represents a significant advancement in healthcare monitoring, offering a responsive and scalable solution to latency and resource constraints in real‐time applications. Future research could explore incorporating additional sensors and advanced ML models to further expand its impact in healthcare applications. [ABSTRACT FROM AUTHOR]

The ubiquitous Internet of Things (IoT) and sensing technologies provide an interesting opportunity of remote health monitoring and disease risk categorisation of populations. An end‐to‐end architecture is proposed to facilitate real‐time digital dashboards to visualise general anxiety risks of patients, especially during a pandemic, such as COVID‐19. To collect physiological data related to anxiety (heart rate, blood pressure, and saturation of peripheral oxygen [SPO2]) and communicate them to a centralised dashboard, dubbed 'X‐DASH', a hardware prototype of the proposed architecture was developed using Node‐MCU and diverse sensors. The dashboard presents a smart categorisation of users' data, assessing their anxiety risks, to provide medical professionals and state authorities a clear visualisation of health risks in populations belonging to different regions. We categorised the risk levels as Normal, Mild, Moderate, Elevated, Severe, and Extreme, based on the collected physiological data and pre‐defined threshold values. The developed hardware prototype in this work was used to collect data from about 500 patients present at cardiac clinic of a leading general hospital in Karachi (Pakistan) and the anxiety risk levels were assigned based on pre‐defined threshold values. To validate the reliability of the X‐DASH, the personal physician of each patient was consulted and was requested to identify each of their anxiety risk levels. It was found that the risk levels suggested by X‐DASH, (based on data of heart rate, blood pressure, and SPO2 were more than 90% accurate when compared with diagnoses of physicians. Subsequently, packet loss, delay and network overhead for the platform was compared when using MQTT, CoAP and Modbus. Although MQTT has shown higher delays, but it is still recommended due to having a higher reliability. [ABSTRACT FROM AUTHOR]

Machine learning methods have been used to solve real-time scheduling problems but none has yet made an architecture that utilizes influences between real-time tasks as input features. This letter proposes a novel approach to partitioned scheduling in real-time systems using graph machine learning. We present a graph representation of real-time task sets that enable graph machine-learning schemes to capture the influence between real-time tasks. By using a graph attention network (GAT) with this method, our model successfully partitioned-schedule task sets that were previously deemed unschedulable by state-of-the-art partitioned scheduling algorithms. The GAT is used to establish relationships between nodes in the graph, which represent real-time tasks, and to learn how these relationships affect the schedulability of the system. [ABSTRACT FROM AUTHOR]

The current quality assessment for fruits and vegetables relies on subjective human judgment and manual inspection, resulting in inconsistencies and inefficiencies. Due to that, there is a need for a real-time system that can accurately and efficiently assess the quality of fruits and vegetables by analyzing various parameters, such as color, texture, size, and blemishes, to ensure consistency and reduce waste in the food supply chain. This study presents the development of a real-time tomato classification system using a hybrid model that combines convolutional neural network (CNN) and support vector machines (SVMs) deployed on the embedded single-board NVIDIA Jetson TX1. The selected CNN model EfficientNetB0 was used for feature extraction and SVM for classification. Notably, the EfficientNetB0-SVM hybrid model demonstrated impressive efficiency, achieving an average accuracy of 93.54% for classifying static tomato images stored in a board into healthy or reject with a testing time of 0.0216-s per image. Also, during real-time implementation, the proposed hybrid model attained an average inference speed of 15.6 frames per second (15.6 FPS), with an accuracy of 78.57% in classifying actual tomatoes into healthy or reject. The classification decision was taken based on 5 images for each tomato captured at different angles to ensure the detection of any blemishes from almost all sides of the tomato. The performance of the proposed model outperforms that of the state-of-the-art (SOTA) methods in accuracy, testing time per image, and real-time prediction accuracy. [ABSTRACT FROM AUTHOR]

Efficiently deploying deep neural networks on resource-limited embedded systems is crucial to meet real-time and power consumption requirements. Utilizing metareasoning as a higher-level controller along with tiny machine learning (TinyML) can enhance energy efficiency and reduce latency on such systems by overseeing available resources. This study introduces MetaTinyML, a comprehensive metareasoning framework for self-guided navigation on TinyML platforms. The framework adapts its decision-making process by factoring in environmental changes to select the most suitable algorithms for the current scenario. Implementation of MetaTinyML on an NVIDIA Jetson Nano 4-GB system integrated with a Jetbot ground vehicle demonstrated up to 50% power consumption enhancement. View a video demonstration of the MetaTinyML framework at: Video. [ABSTRACT FROM AUTHOR]

Perception-based autonomous system design methods are widely adopted in various domains like transportation, industrial robotics, etc. However, attaining safe and predictable execution in such systems depends on the platform-level integration of perception and control tasks. This letter presents a novel methodology to co-optimize these tasks, assuming a CPU-GPU-based real-time platform, a common choice of compute resource in this domain. Unlike the traditional methods that separately address AI-based sensing and control concerns, we consider that the overall performance of the system depends on the inferencing accuracy of the perception tasks and the performance of the control tasks iteratively executing in a feedback loop. We propose a design-space exploration methodology that considers the above concern and validates the same on an autonomous driving use case using a novel simulation setup. [ABSTRACT FROM AUTHOR]

As the demand for edge computing applications continues to rise, the need for efficient training of resource-constrained devices becomes paramount. This letter proposes message queuing telemetry transport (MQTT)-based implementation of distributed estimation strategies in the context of the Internet of Things (IoT), namely incremental, consensus, and diffusion strategies. The use of Raspberry Pi Pico W in the emulation environment is motivated by its advanced capability, while the MQTT data protocol is employed to address the constraints associated with conventional HTTP/HTTPs protocols. Synchronization in an IoT network is achieved by the integration of a novel methodology that entails the use of the wait-for-slowest (WFS) protocol and the MQTT protocol. Furthermore, the development of a graphical user interface supported by the Django application allows for adjusting parameters in distributed strategies through the HTTP REST API, along with SQLite. The results acquired from hardware experiments exhibit a strong correlation between the mean-square performance achieved from simulation studies. The distributed estimation strategy is compared with state-of-the art centralized and noncooperation estimation strategies, demonstrating its superior performance. In addition, a study is conducted on the resilience of these IoT networks in the face of several network threats, such as node failure and model poisoning attacks. A theoretical analysis is provided to explain the relationship between the number of iterations and node failure. [ABSTRACT FROM AUTHOR]

Ensuring the safety of water airport runways is essential for the correct operation of seaplane flights. Among other tasks, airport operators must identify and remove various objects that may have drifted into the runway area. In this paper, the authors propose a complete and embedded‐friendly waterway obstacle detection pipeline that runs on a camera‐equipped drone. This system uses a class‐agnostic version of the YOLOv7 detector, which is capable of detecting objects regardless of its class. Additionally, through the usage of the GPS data of the drone and camera parameters, the location of the objects are pinpointed with 0.58 m Distance Root Mean Square. In our own annotated dataset, the system is capable of generating alerts for detected objects with a recall of 0.833 and a precision of 1. [ABSTRACT FROM AUTHOR]

Generalized frequency division multiplexing (GFDM) is a flexible block‐structured multi‐carrier scheme recently proposed for next‐generation wireless communication systems. There are various approaches suggested for its analysis and implementation via simulations but testing in real‐time environments is not heavily investigated. This paper carries out the real‐time implementation of the GFDM system utilizing software‐defined radio (SDR) by emphasizing mainly channel estimation and synchronization. Symbol timing, frequency offset, and channel estimate algorithms are applied using a windowed preamble with two identical halves to satisfy low egress noise requirements. Time and frequency estimation is evaluated in terms of residual offsets along with symbol error rate over frequency selective channels. This algorithm is extended to a preamble composed of multiple identical parts. This facilitates a large frequency estimation range at the cost of complexity. For practical validation of the above concepts, the National Instruments (NI) universal software radio peripheral (USRP) 2953R is employed as hardware and it is interfaced with LabVIEW. [ABSTRACT FROM AUTHOR]

This paper presents a hardware implementation of a one-dimensional convolutional neural network using depthwise separable convolution (DSC) on the VC707 FPGA development board. The design processes the one-dimensional rolling bearing current signal dataset provided by Paderborn University (PU), employing minimal preprocessing to maximize the comprehensiveness of feature extraction. To address the high parameter demands commonly associated with convolutional neural networks (CNNs), the model incorporates DSC, significantly reducing computational complexity and parameter load. Additionally, the DoReFa-Net quantization method is applied to compress network parameters and activation function outputs, thereby minimizing memory usage. The quantized DSC model requires approximately 22 KB of storage and performs 1,203,128 floating-point operations in total. The implementation achieves a power consumption of 527 mW at a clock frequency of 50 MHz, while delivering a fault diagnosis accuracy of 96.12%. [ABSTRACT FROM AUTHOR]

Advancements in neural network approaches have enhanced the effectiveness of surface Electromyography (sEMG)-based hand gesture recognition when measuring muscle activity. However, current deep learning architectures struggle to achieve good generalization and robustness, often demanding significant computational resources. The goal of this paper was to develop a robust model that can quickly adapt to new users using Transfer Learning. We propose a Multi-Scale Convolutional Neural Network (MSCNN), pre-trained with various strategies to improve inter-subject generalization. These strategies include domain adaptation with a gradient-reversal layer and self-supervision using triplet margin loss. We evaluated these approaches on several benchmark datasets, specifically the NinaPro databases. This study also compared two different Transfer Learning frameworks designed for user-dependent fine-tuning. The second Transfer Learning framework achieved a 97% F1 Score across 14 classes with an average of 1.40 epochs, suggesting potential for on-site model retraining in cases of performance degradation over time. The findings highlight the effectiveness of Transfer Learning in creating adaptive, user-specific models for sEMG-based prosthetic hands. Moreover, the study examined the impacts of rectification and window length, with a focus on real-time accessible normalizing techniques, suggesting significant improvements in usability and performance. [ABSTRACT FROM AUTHOR]

Traffic congestion is a problem that burdens urban cities daily. Despite the development of numerous research methodologies and technological solutions, the problem still persists. This article presents a system that combines traffic and demand management mechanisms to avoid the emergence of congestion by sustaining the occupancy of each road in the network up to a threshold, using a reservation scheme and keeping track of the future states of the network. Traffic management is responsible for navigating vehicles through congestion-free paths, while demand management is responsible for identifying the time when each vehicle will enter the network. Considering the size of actual urban networks and the number of vehicles utilizing the infrastructure, this is not an easy task. This work designs and validates a responsive and scalable real-life system for online traffic and demand management, while addressing the challenge of managing and processing large numbers of requests and data. The complexity and feasibility of the system are evaluated through microsimulations of real and artificial road networks. Its traffic efficiency is also evaluated by running extensive simulations of a real city, emulating realistic conditions in different scenarios. [ABSTRACT FROM AUTHOR]

Unmanned aerial vehicles (UAV) can capture multiview images of large-scale scenes, and then using aerial light field (ALF)-rendering technology, they can render high-quality, large-scale 3D scenes. However, the reconstruction of large-scale scenes poses challenges, such as rendering distortion and high memory consumption. In this article, we study the multiview image-capturing and novel view-rendering method of UAV sampling to address these issues. First, we present an ALF sampling model using the spectral analysis of light field and obtain an exact spectrum expression of the ALF. Through the spectral support of ALF, we determine the bandwidth and calculate the minimum required UAV sampling rate. Finally, we demonstrate that our sampling and rendering methods can improve the rendering quality of UAV 3D reconstruction and reduce the minimum sampling rate. [ABSTRACT FROM AUTHOR]

Visual analytics is vital for identifying targets by differentiating their structure and texture from unmanned aerial vehicles (UAVs). Object sensing disturbance and swift texture differentiation are tedious due to the UAV displacements. For improving the accuracy of target detection, this article introduces a learning optimizer-based visual analytics method. The proposed method assimilates deep learning and a gradient descent algorithm for feature differentiation and error minimization concurrently. The captured images are identified using multiple structural feature variations and are correlated with similar stored images. The features are extracted at different displacement and structural changing instances for leveraging accuracy. The learning process trains the similarity features during different differentiation factors. In the feature extraction, the minimum slope points are identified using a gradient descent algorithm by assigning random weights. As the differentiation increases using similar features, the minimum similarity value is detection. Postdetection, the weights are incremental and linear across different feature slopes. Therefore, the accuracy increases under varying displacement instances, preventing target misdetection. The gradient function is invariable between the minimum and maximum values for identifying high-precision features. This ensures optimal detection of different buildings and structures with high accuracy. [ABSTRACT FROM AUTHOR]

This paper presents an innovative scheduling algorithm designed specifically for real-time energy harvesting systems, with a primary focus on minimizing energy consumption and extending the battery's lifespan. The algorithm employs a fixed priority assignment which is the deadline monotonic policy, we have chosen it for its optimality and superior performance compared to other fixed priority scheduling methods. To achieve a balance between energy efficiency and system performance, we incorporated a DVFS (Dynamic Voltage and Frequency Scaling) technique into the algorithm. This adaptive approach enables precise control over the processor's operating frequency, effectively managing energy consumption while ensuring satisfactory system functionality. The core objective of our scheduling algorithm centers on optimizing energy utilization in real-time energy harvesting systems, specifically tailored to extend the battery's operational life. Rigorous evaluations, including comprehensive comparisons against established fixed priority scheduling algorithms, validate the algorithm's efficacy in significantly reducing energy consumption while preserving the system's overall functionality. By combining the deadline monotonic policy and DVFS technique, our proposed algorithm emerges as a promising solution for energy-autonomous systems, contributing to the advancement of sustainable energy practices in real-time applications. As energy harvesting technologies continue to progress, our algorithm holds valuable potential to provide critical insights for enhancing the efficiency and reliability of future energy harvesting systems. [ABSTRACT FROM AUTHOR]

UPPAAL is a formal modeling and verification tool based on timed automata, capable of effectively analyzing real‐time software and hardware systems. In this article, we investigate research on UPPAAL‐assisted formal modeling and verification. First, we propose four research questions considering tool characteristics, modeling methods, verification means and application domains. Then, the state‐of‐the‐art methods for model specification and verification in UPPAAL are discussed, involving model transformation, model repair, property specification, as well as verification and testing methods. Next, typical application cases of formal modeling and verification assisted by UPPAAL are analyzed, spanning across domains such as network protocol, multi‐agent system, cyber‐physical system, rail traffic and aerospace systems, cloud and edge computing systems, as well as biological and medical systems. Finally, we address the four proposed questions based on our survey and outline future research directions. By responding to these questions, we aim to provide summaries and insights into potential avenues for further exploration in this field. [ABSTRACT FROM AUTHOR]

As real-time systems are safety critical, guaranteeing a high reliability threshold is as important as meeting all deadlines. Periodic tasks are replicated to mitigate the negative impact of transient faults, which leads to redundancy and high energy consumption. On the other hand, energy saving is widely identified as increasingly relevant issues in real-time systems. In this paper, we formalize this challenging tri-criteria optimization problem, i.e., minimizing the expected energy consumption while enforcing the reliability threshold and meeting all task deadlines, and propose several mapping and scheduling heuristics to solve it. Specifically, a novel approach is designed to (i) map an arbitrary number of replicas onto processors, (ii) schedule each replica of each task instance on its assigned processor with less temporal overlap. The platform is composed of processing units with different characteristics, including speed profile, energy cost and fault rate. The heterogeneity of the computing platform makes the problem more complicated, because different mappings achieve different levels of reliability and consume different amounts of energy. Moreover, scheduling plays an important role in energy saving, as the expected energy consumption is the average over all failure scenarios. Once a task replica is successful, the other replicas of that task instance can be canceled, which calls for minimizing the overlap between any replica pair. Finally, to quantitatively analyze our methods, we derive a theoretical lower-bound for the expected energy consumption. Comprehensive experiments are conducted on a large set of execution scenarios and parameters. The comparison results reveal that our strategies perform better than the random baseline under almost all settings, with an average gain in energy consumption of more than 40%, and our best heuristic achieves an excellent performance: its energy saving is only 2% less than the lower-bound on average. [ABSTRACT FROM AUTHOR]

This paper presents a real-time wearable system designed to assist Parkinson's disease patients experiencing freezing of gait episodes. The system utilizes advanced machine learning models, including convolutional and recurrent neural networks, enhanced with past sample data preprocessing to achieve high accuracy, efficiency, and robustness. By continuously monitoring gait patterns, the system provides timely interventions, improving mobility and reducing the impact of freezing episodes. This paper explores the implementation of a CNN+RNN+PS machine learning model on a microcontroller-based device. The device operates at a real-time processing rate of 40 Hz and is deployed in practical settings to provide 'on demand' vibratory stimulation to patients. This paper examines the system's ability to operate with minimal latency, achieving an average detection delay of just 261 milliseconds and a freezing of gait detection accuracy of 95.1%. While patients received on-demand stimulation, the system's effectiveness was assessed by decreasing the average duration of freezing of gait episodes by 45%. These preliminarily results underscore the potential of personalized, real-time feedback systems in enhancing the quality of life and rehabilitation outcomes for patients with movement disorders. [ABSTRACT FROM AUTHOR]

The integration of cloud computing (CC) and Internet of Things (IoT) technologies in the healthcare industry has significantly boosted the importance of real-time remote patient monitoring. The Internet of Medical Things (IoMT) systems facilitate the seamless transfer of health records to data centers, allowing medical professionals and caregivers to analyze, process, and access them. This data is often stored in cloud-based systems. Nevertheless, the transmission of data and execution of computations in a cloud environment may lead to delays and affect the efficiency of real-time healthcare services. In addition, the use of edge computing (EC) layers has become prevalent in performing local data processing and storage to reduce service response times for IoMT applications. The main objective of this article is to develop an adaptive EC infrastructure for IoMT systems, with a specific emphasis on maintaining optimal performance for real-time health services. It also designs a model to predict the server resources required to meet service level agreements (SLAs) regarding response time. Simulation results demonstrate that EC significantly improves service response time for real-time IoMT applications. The proposed model can accurately and efficiently predict the computing resources required for medical data services to achieve SLAs under varying workload conditions. [ABSTRACT FROM AUTHOR]

The diffusion of electric vehicles (EVs) is recently gaining great attention in the road transport and automotive sectors as an attempt to bring in an emission-free world. EVs are considered a key to future clean transportation systems. However, these vehicles still suffer from limited battery capacity and range anxiety. Therefore, EVs manufacturers are focusing on reducing energy consumption and CO2 emissions. In addition, research in the context of intelligent transportation systems embedding information and communication technologies are focusing on the optimization of the energy consumption as a valuable solution to foster the wide diffusion of EVs. In this article, we propose a simulation platform for eco-routing services based on estimating EV energy consumption to provide the most energy-efficient routes for the EV while traveling. We provide an energy map that can be used for eco-routing through a real-time data collection of the EV energy consumption. The energy map was established in the traffic simulator Simulation of Urban MObility (SUMO) to show the efficiency of the proposed eco-routing strategy compared to the other strategies based on establishing the fastest routes. This map will be exploited as good support, in the future, for advanced research on the EV concept. [ABSTRACT FROM AUTHOR]

IEEE 802.1 Time‐Sensitive Networking (TSN) has proposed several shapers, for example, time‐aware shaper (TAS, 802.1Qbv), asynchronous traffic shaping (ATS, 802.1Qcr), credit‐based shaper (CBS, 802.1Qav), and cyclic queuing and forwarding (CQF, 802.1Qch). The shapers have their advantages and disadvantages and can be used in isolation or in combination to address the varied timing requirements of industrial application streams. There is very limited work on how to analyse and configure shaper combinations. The authors are interested in the configuration optimisation of multi‐shaper TSN networks, targeting the TAS + CBS, TAS + ATS, and TAS + Multi‐CQF combinations. The authors first propose multi‐shaper integration approaches, focusing on a novel iterative delay analysis for TAS + ATS, an approach to integrate TAS and CQF by placing constraints on TAS scheduling as well as the TAS and CBS integration. We formulate the combinatorial optimisation problem of configuring multi‐shaper TSN networks, which consists, for example, of the routing of streams, the assignment of streams to the egress port queues, and the synthesis of gate control lists for TAS. Then, the authors propose a solution based on a simulated annealing metaheuristic. The proposed solutions are evaluated on large realistic test cases, up to tens of thousands of streams and devices. [ABSTRACT FROM AUTHOR]

Real-time Hand Gesture Recognition (HGR) has emerged as a vital technology in human-computer interaction, offering intuitive and natural ways for users to interact with computer-vision systems. This comprehensive review explores the advancements, challenges, and future directions in real-time HGR. Various HGR-related technologies have also been investigated, including sensors and vision technologies, which are utilized as a preliminary step in acquiring data in HGR systems. This paper discusses different recognition approaches, from traditional handcrafted feature methods to state-of-the-art deep learning techniques. Learning paradigms have been analyzed such as supervised, unsupervised, transfer, and adaptive learning in the context of HGR. A wide range of applications has been covered, from sign language recognition to healthcare and security systems. Despite significant developments in the computer vision domain, challenges remain in areas such as environmental robustness, gesture complexity, computational efficiency, and user adaptability. Lastly, this paper concludes by highlighting potential solutions and future research directions trying to develop more robust, efficient, and user-friendly real-time HGR systems. [ABSTRACT FROM AUTHOR]

In real-time systems analysis, probabilistic models, particularly Markov chains, have proven effective for tasks with dependent executions. This paper improves upon an approach utilizing Gaussian emission distributions within a Markov task execution model that analyzes bounds on deadline miss probabilities for tasks in a reservation-based server. Our method distinctly addresses the issue of runtime complexity, prevalent in existing methods, by employing a state merging technique. This not only maintains computational efficiency but also retains the accuracy of the deadline-miss probability estimations to a significant degree. The efficacy of this approach is demonstrated through the timing behavior analysis of a Kalman filter controlling a Furuta pendulum, comparing the derived deadline miss probability bounds against various benchmarks, including real-time Linux server metrics. Our results confirm that the proposed method effectively upper-bounds the actual deadline miss probabilities, showcasing a significant improvement in computational efficiency without significantly sacrificing accuracy. [ABSTRACT FROM AUTHOR]

The development of static real-time control systems often follows a closed-world assumption, allowing extensive RTOS-aware whole-program optimization. For single-core systems, previous work could show the high potential of control-flow-aware static system-call tailoring. However, due to an exponential state explosion in the analysis phase, it cannot simply be extended to a multi-core setting, since the core's relative timing to each other is undetermined. In this work, we present MultiSSE, a multi-core capable and RTOS-aware static whole-system optimization. First, MultiSSE analyzes the system by determining the relative positions of multiple cores only when necessary. For that, it exploits structural control flow and optionally timing information to handle each core separately as much as possible. Based on the analysis result, a synthesis applies lock elision and system-call optimization to generate specialized multi-core real-time systems for AUTOSAR OS. To enable a static prediction of the run-time reduction, we additionally provide cost models for the optimized cross-core system calls and evaluate the approach with synthetic benchmarks and a real-world quadrotor application. MultiSSE was able to optimize or even completely elide costly cross-core system calls and system objects leading to a reduction of up to 14% of a task's execution time. In this extended version of a conference publication (Entrup et al. 2023), we provide an advanced description, new cost models, and an end-to-end measurement by developing a synthesis complementing the analysis. [ABSTRACT FROM AUTHOR]

An online non‐model‐based procedure is presented for estimating the synchronous generator (SG) dynamic parameters using practical phasor measurement unit (PMU) signals in the presence of uncertainty and noisy data. For this purpose, considering 8th‐order approximation, the SGs model is estimated in which, based on evaluating voltage and current phasors achieved from PMU data, dynamic parameters are estimated online. The proposed approach is a generalised concept of the Heffron–Philips model in which the variables and the gain factors are adaptable according to operating conditions. The proposed scheme is an online and non‐model‐based method in which the SG magnetic saturation behaviours are modelled through multivariable non‐linear definition to extend the accurate controlling structure. In this case, two different studies are carried out. In the first study, considering a single SG is connected to the infinite bus, the ability of the proposed method through simulation studies is evaluated. In the second study, the proposed scheme is developed practically in the laboratory whereby performing the experimental structure on different types through real‐time working mode, validation of the proposed estimated model through different operating points is evaluated. Experimental results show the effectiveness of the proposed practical scheme for estimating the generator's detailed model and non‐linear dynamic parameters through real‐time evaluations. [ABSTRACT FROM AUTHOR]

For flight test engineering, the flight test task schedule is of great importance to the delivery node and the development cost of an aircraft, while in the real flight test process, dynamic events may frequently occur, which affect the schedule implementation and flight test progress. To adaptively adjust the real-world flight test schedule, this paper proposes a deep reinforcement learning (DRL) approach to solve the dynamic task scheduling problem for flight tests, with the objectives of flight test duration and task tardiness. Firstly, the task scheduling characteristics for flight tests are introduced, and a mixed-integer programming (MIP) model is constructed. Then, the addressed problem is formulated as a Markov decision process (MDP), including the well-designed state features, reward functions, and action space based on the heuristic rules for selecting the uncompleted flight test task and allocating the selected task to an appropriate aircraft. Proximal policy optimization (PPO) is adopted to train and learn the optimal policy. Finally, extensive experiments are carried out to verify the proposed method's effectiveness and efficiency in constructing a high-quality flight test task schedule in a dynamic flight test environment. [ABSTRACT FROM AUTHOR]

This letter introduces the design of an estimator for parameters of Middleton Class A noise using its canonical formula and classical numerical methods. The main focus is to acquire parameters to characterize communication channels in intelligent systems or those based on cognitive paradigms. A comprehensive analysis of the first-order characteristics of the Middleton Class A noise model is conducted to establish the foundational understanding necessary for developing the presented estimator model, named impulsive noise estimator with minimization methods (INEMM). Subsequently, the method is introduced, substantiated, and compared to various established estimators concerning precision and complexity. Results show a distinct advantage in terms of overall performance. [ABSTRACT FROM AUTHOR]

In the literature, macroscopic fundamental diagram (MFD)- or network fundamental diagram-based perimeter control approaches have been demonstrated as cost-effective means to improve traffic efficiency for urban traffic networks. This article proposes a new perimeter control approach, which extends previous approaches in two aspects. First, the proposed approach optimizes the allocation of perimeter metered flow considering the heterogeneity of network traffic flow distribution, which is inversely related to network production. Second, queue management is considered by the proposed approach so as to prevent queue spillback induced by perimeter control. The proposed control approach is well advanced for field application using only loop detector data. It adopts a hierarchical framework, where the upper level controller determines the total flow entering the protected network through perimeter links based on an MFD-based feedback regulator, while the lower level controller optimizes the distribution of the total metered flow among the perimeter links based on real-time network heterogeneity and perimeter queues. The proposed approach is tested and compared with state-of-the-art perimeter control approaches, using a real traffic network developed in the microscopic simulation software VISSIM. Results demonstrate that the proposed approach significantly reduces the overall network delay. Moreover, it outperforms other perimeter control approaches by maintaining low network heterogeneity within the protected network and preventing (or delaying) queue spillback at perimeter links. [ABSTRACT FROM AUTHOR]

Purpose: Wearable ultrasound devices can be used to continuously monitor muscle activity. One possible application is to provide real-time feedback during physiotherapy, to show a patient whether an exercise is performed correctly. Algorithms which automatically analyze the data can be of importance to overcome the need for manual assessment and annotations and speed up evaluations especially when considering real-time video sequences. They even could be used to present feedback in an understandable manner to patients in a home-use scenario. The following work investigates three deep learning based segmentation approaches for abdominal muscles in ultrasound videos during a segmental stabilizing exercise. The segmentations are used to automatically classify the contraction state of the muscles. Methods: The first approach employs a simple 2D network, while the remaining two integrate the time information from the videos either via additional tracking or directly into the network architecture. The contraction state is determined by comparing measures such as muscle thickness and center of mass between rest and exercise. A retrospective analysis is conducted but also a real-time scenario is simulated, where classification is performed during exercise. Results: Using the proposed segmentation algorithms, 71% of the muscle states are classified correctly in the retrospective analysis in comparison to 90% accuracy with manual reference segmentation. For the real-time approach the majority of given feedback during exercise is correct when the retrospective analysis had come to the correct result, too. Conclusion: Both retrospective and real-time analysis prove to be feasible. While no substantial differences between the algorithms were observed regarding classification, the networks incorporating the time information showed temporally more consistent segmentations. Limitations of the approaches as well as reasons for failing cases in segmentation, classification and real-time assessment are discussed and requirements regarding image quality and hardware design are derived. [ABSTRACT FROM AUTHOR]

This paper presents an energy-efficient and high-accuracy sampling synchronization approach for real-time synchronous data acquisition in wireless sensor networks (saWSNs). A proprietary protocol based on time-division multiple access (TDMA) and deep energy-efficient coding in sensor firmware is proposed. A real saWSN model based on 2.4 GHz nRF52832 system-on-chip (SoC) sensors was designed and experimentally tested. The obtained results confirmed significant improvements in data synchronization accuracy (even by several times) and power consumption (even by a hundred times) compared to other recently reported studies. The results demonstrated a sampling synchronization accuracy of 0.8 μs and ultra-low power consumption of 15 μW per 1 kb/s throughput for data. The protocol was well designed, stable, and importantly, lightweight. The complexity and computational performance of the proposed scheme were small. The CPU load for the proposed solution was <2% for a sampling event handler below 200 Hz. Furthermore, the transmission reliability was high with a packet error rate (PER) not exceeding 0.18% for TXPWR ≥ −4 dBm and 0.03% for TXPWR ≥ 3 dBm. The efficiency of the proposed protocol was compared with other solutions presented in the manuscript. While the number of new proposals is large, the technical advantage of our solution is significant. [ABSTRACT FROM AUTHOR]

Emotion recognition systems are typically trained to classify a given psychophysiological state into emotion categories. Current platforms for emotion ground-truth collection show limitations for real-world scenarios of long-duration content (e.g. $> $ > 10 minutes), namely: 1) Real-time annotation tools are distracting and become exhausting; 2) Perform retrospective annotation of the whole content in bulk (providing highly coarse annotations); or 3) Are used by external experts (depending on the number of annotators and their subjective experience). We explore a novel approach, the EmotiphAI Annotator, that allows undisturbed content visualisation and simplifies the annotation process by using segmentation algorithms that select brief clips for emotional annotation retrospectively. We compare three methods for content segmentation based on physiological data (Electrodermal Activity (EDA), emotion-based), scene (time-based), and random (control) selection. The EmotiphAI Annotator attained a B+ System Usability Scale score and low-average mental workload as per the NASA Task Load Index (40%). The reliability of the self-report was analysed by the inter-rater agreement (STD $< $ < 0.75), coherence across time segmentation methods (STD $< $ < 0.17), comparison against the state-of-the-art ground truth (STD $< $ < 0.7), and correlation to EDA ($> $ > 0.3 to 0.8), where the EDA-based method obtained the overall best performance. [ABSTRACT FROM AUTHOR]

Emotions play an essential role in affecting the performance of cognitive abilities in continuous cognitive tasks. Most previous studies share a common issue in that the evoked emotions are simply presumed to be real emotions, without taking into account the observation that emotions may be changed when carrying out cognitive activities. This may lead to the inaccurate detection of true emotions, which further adversely affects the investigation of interactions between emotion and cognition. To address this challenging problem, the present work develops an innovative study using EEG measurement to investigate the interaction between dynamic affection and cognitive ability. In particular, a real-time emotion detection model by the use of physiological signals (i.e., EEG) is constructed, to dynamically monitor the current emotional state. Given the observed emotion, the analysis of the interaction between cognitive abilities and dynamic emotions is undertaken from the perspectives of both behavioral performance and brain mechanisms. Research outcomes indicate that emotions are not stable, and are indeed dynamically changed by cognitive performance. Meanwhile, cognitive activities also influence the brain activation pattern revealed under different emotions, which validates the necessity of introducing the dynamic emotion monitoring model. In addition, the best performance has been found when the emotional state is neutral in terms of accuracy and response time. The results of this study provide a potential basis for assessing the cognitive abilities of individuals with different emotions in a variety of applications of cognitive scenarios. [ABSTRACT FROM AUTHOR]

An affective brain-computer interface (aBCI) can detect affective states based on brain signals and might assist people in improving their emotion regulation abilities. However, individual differences in emotional brain patterns make cross-subject emotion identification extremely challenging. Traditional supervised single-subject classification schemes require considerable calibration samples from new individuals to train subject-dependent models. Individuals are easily fatigued with long-term EEG collection processes, which may affect performance in subsequent online experiments. In this study, we propose a real-time aBCI system using domain-fusion-based multisource style transfer mapping (DF-MS-STM) to detect positive, neutral, and negative emotional states without the need for additional training sessions. Sixteen subjects participated in our online experiments to test the performance of our aBCI system and an average online prediction accuracy of 72.17±12.25% was obtained for three-class emotion recognition tasks in the last three experimental sessions. Our proposed algorithm significantly outperformed numerous baseline methods in terms of cross-subject emotion classification. In addition, we identified distinct brain patterns in response to different emotional stimuli based on the results of event-related spectral perturbation (ERSP) analyses. These neural patterns might provide new insights for emotional brain mechanistic studies and related aBCIs. [ABSTRACT FROM AUTHOR]

The recent shift in the VLSI industry from conventional MOSFET to FinFET for designing contemporary chip-multiprocessor (CMP) has noticeably improved hardware platforms' computing capabilities, but at the cost of several thermal issues. Unlike the conventional MOSFET, FinFET devices experience a significant increase in circuit speed at a higher temperature, called temperature effect inversion (TEI), but higher temperature can also curtail the circuit lifetime due to self-heating effects (SHEs). These fundamental thermal properties of FinFET introduced a new challenge for scheduling time-critical tasks on FinFET-based multicores that how to exploit TEI towards improving performance while combating SHEs. In this work, TREAFET, a temperature-aware real-time scheduler, attempts to exploit the TEI feature of FinFET-based multicores in a time-critical computing paradigm. At first, the overall progress of individual tasks is monitored, tasks are allocated to the cores, and finally, a schedule is prepared. By considering the thermal profiles of the individual tasks and the current thermal status of the cores, hot tasks are assigned to the cold cores and vice-versa. Finally, the performance and temperature are balanced on-the-fly by incorporating a prudential voltage scaling towards exploiting TEI while guaranteeing the deadline and thermal safety. Moreover, TREAFET stimulates the average runtime frequency by employing an opportunistic energy-adaptive voltage spiking mechanism, in which energy saving during memory stalls at the cores is traded off during the time slice having the spiked voltage. Simulation results claim TREAFET maintains a safe and stable thermal status (peak temperature below 80 °C) and improves frequency up to 17% over the assigned value, which ensures legitimate time-critical performance for a variety of workloads while surpassing a state-of-the-art technique. The stimulated frequency in TREAFET also finishes the tasks early, thus providing opportunities to save energy by power gating the cores, and achieves a 24% energy delay product (EDP) gain on average. [ABSTRACT FROM AUTHOR]

The automotive industry has recently adopted multicore processors and microcontrollers to meet the requirements of new features, such as autonomous driving, and comply with the latest safety standards. However, inter-core communication poses a challenge in ensuring real-time requirements such as time determinism and low latencies. Concurrent access to shared buffers makes predicting the flow of data difficult, leading to decreased algorithm performance. This study explores the integration of Logical Execution Time (LET) and Time-Division Multiple Access (TDMA) models in multicore embedded systems to address the challenges in inter-core communication by synchronizing read/write operations across different cores, significantly reducing latency variability and improving system predictability and consistency. Experimental results demonstrate that this integrated approach eliminates data loss and maintains fixed operation rates, achieving a consistent latency of 11 ms. The LET-TDMA method reduces latency variability to approximately 1 ms, maintaining a maximum delay of 1.002 ms and a minimum delay of 1.001 ms, compared to the variability in the LET-only method, which ranged from 3.2846 ms to 8.9257 ms for different configurations. [ABSTRACT FROM AUTHOR]

Copyright of Technology Audit & Production Reserves is the property of PC TECHNOLOGY CENTER and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

This research presents an autonomous Real-Time Facial Recognition Attendance System (FRAS) that automates and refines traditional attendance tracking. Utilizing advanced facial recognition algorithms, FRAS eliminates manual marking by leveraging a self-made dataset adaptable to diverse faces and settings. The system captures live video, detects faces, and identifies individuals instantly, updating attendance records in real-time for a secure, efficient, and adaptable solution compared to manual methods. [ABSTRACT FROM AUTHOR]

Addressing the escalating necessity for proficient and dependable autonomous vehicle systems, this research proposes an advanced object detection approach harnessing the YOLOv8 algorithm. Object detection plays a pivotal role in the functionality of autonomous driving systems, enabling vehicles to perceive and react to their surroundings promptly. YOLOv8, an upgraded iteration of the YOLO algorithm, is acclaimed for its rapidity and precision in object detection endeavors. Our proposed model, distinguished by enhancements in network architecture and training methodologies, surpasses existing models in terms of detection precision and computational efficacy. Through an exhaustive examination, we delve into the network architecture, training regimen, and assessment metrics of the YOLOv8-based model. Experimental findings showcase the model's commendable performance in both quantitative and qualitative assessments, highlighting its resilience in identifying pedestrians, vehicles, traffic signs, and other pertinent objects across varied driving scenarios. The model's predicted outputs attest to its adeptness in precisely localizing and categorizing objects of interest, thereby augmenting the safety and efficiency of autonomous driving systems. [ABSTRACT FROM AUTHOR]

To validate safety-related automotive software systems, experimental tests are conducted at different stages of the V-model, which are referred as "X-in-the-loop (XIL) methods". However, these methods have significant drawbacks in terms of cost, time, effort and effectiveness. In this study, based on hardware-in-the-loop (HIL) simulation and real-time fault injection (FI), a novel testing framework has been developed to validate system performance under critical abnormal situations during the development process. The developed framework provides an approach for the real-time analysis of system behavior under single and simultaneous sensor/actuator-related faults during virtual test drives without modeling effort for fault mode simulations. Unlike traditional methods, the faults are injected programmatically and the system architecture is ensured without modification to meet the real-time constraints. Moreover, a virtual environment is modeled with various environmental conditions, such as weather, traffic and roads. The validation results demonstrate the effectiveness of the proposed framework in a variety of driving scenarios. The evaluation results demonstrate that the system behavior via HIL simulation has a high accuracy compared to the non-real-time simulation method with an average relative error of 2.52. The comparative study with the state-of-the-art methods indicates that the proposed approach exhibits superior accuracy and capability. This, in turn, provides a safe, reliable and realistic environment for the real-time validation of complex automotive systems at a low cost, with minimal time and effort. [ABSTRACT FROM AUTHOR]

The advent of 6G communication systems has enabled, through integrated sensing and communications (ISAC), the vision of the intelligent connection of everything. Since such ISAC systems will be utilizing the sub-6-GHz and millimeter-wave (mm-wave) bands, appropriate hybrid, i.e., sensing and communication channel models, are required for their proper operation. This article first presents a comprehensive survey of channel models used in ISAC applications viewed from different modeling perspectives by identifying the challenges encountered in these modeling approaches. A novel hybrid approach for ISAC channel modeling is then introduced wherein the modeling methodology is partitioned into three elements: targets, clutters, and interferences. The advantage of such a flexible modeling approach using these three elements enhances the proposed model’s ability to effectively deal with the specific sensing and communication operational requirements. Finally, further performance enhancement methods, based upon real-time 3D environment reconstruction techniques as well as artificial intelligence (AI), are proposed and discussed. [ABSTRACT FROM AUTHOR]

A 1 millisecond (1‐ms) vision system that guarantees high efficiency and timely response for tomato defect detection is essential for factory automation. Because of various defect appearances, recently many existing researches focus on CNN based defect detection, but few of them attempt to reach high processing speed to adapt to the factorial assembly line. This paper proposes a global to multi‐scale local based parallel architecture with hardwired CNN for tomato defect detection. This architecture breaks down image‐wise detection into pixel‐wise localization and block‐wise classification. The pixel‐wise localization utilizes tomato‐aware information as constraints for localization performance. The block‐wise classification uses a fully pipelined network structure to obtain the classification result for each block as the pixel stream moves through the network. The classification network has a six‐layer lightweight network structure with quantization for hardwired type implementation on FPGA. The experiment results show that the proposed architecture processes 1000 FPS images with 0.9476 ms/frame delay. And for detection performance, this architecture keeps f1−score$f_{1}-score$ at 80.18%, only 1.31% lower than ResNet50 based detection system. [ABSTRACT FROM AUTHOR]