Search

In this paper, we analyze a hybrid Heterogeneous Cellular Network (HCNet) framework by deploying millimeter Wave (mmWave) small cells with coexisting traditional sub-6GHz macro cells to achieve improved coverage and high data rate. We consider randomly-deployed macro base stations throughout the network whereas mmWave Small Base Stations (SBSs) are deployed in the areas with high User Equipment (UE) density. Such user centric deployment of mmWave SBSs inevitably incurs correlation between UE and SBSs. For a realistic scenario where the UEs are distributed according to Poisson cluster process and directional beamforming with line-of-sight and non-line-of-sight transmissions is adopted for mmWave communication. By using tools from stochastic geometry, we develop an analytical framework to analyze various performance metrics in the downlink hybrid HCNets under biased received power association. For UE clustering we considered Thomas cluster process and derive expressions for the association probability, coverage probability, area spectral efficiency, and energy efficiency. We also provide Monte Carlo simulation results to validate the accuracy of the derived expressions. Furthermore, we analyze the impact of mmWave operating frequency, antenna gain, small cell biasing, and BSs density to get useful engineering insights into the performance of hybrid mmWave HCNets. Our results show that network performance is significantly improved by deploying millimeter wave SBS instead of microwave BS in hot spots.

A novel channel estimation method based on deep learning algorithm is proposed for large-scale IoT networks. We consider asymmetric backscatter communication system to maintain low-power at sensor nodes. In order to obtain channel data, we design denoising autoencoder which consists of encoder with Feedforward Neural Network (FNN) and decoder with Convolutional Neural Network (CNN). Finally, the channel estimation error is minimized, while the pilots are optimized. Especially, we adopt beamforming technique that relies only on cascaded channel data to reduce complexity in multi-sensor system. It is shown that the accuracy is slightly degraded while the complexity is greatly reduced.

Abstract In the domain of modern manufacturing, computer numerical control (CNC) milling machines have emerged as instrumental assets. However, the data they generate is of vast amount, but usually contains redundancies and displays consistent patterns, making it inefficient for deep learning training. This paper proposes a novel sampling algorithm tailored for CNC milling machine data, emphasizing both diversity and efficiency. The proposed method leverages the entropy concept from the information‐theoretic perspective to evaluate and enhance data diversity, aiming to achieve efficient learning with high accuracy. This in turn enables to not only facilitates a deeper understanding of CNC data characteristics but also contributes significantly to the optimization of deep learning training processes in the context of CNC milling data.

In order to examine the potential and synergetic aspects of intelligent reflecting surface (IRS) techniques for Internet-of-Things (IoT), we study an IRS-aided Long Range (LoRa) system in this paper. Specifically, to facilitate the acquisition of accurate channel state information (CSI) for effective reflection of LoRa signals, we first propose an optimal training design for the least squares channel estimation with LoRa modulation, and then, by utilizing the acquired CSI, we develop a high-performing passive beamforming scheme based on a signal-to-ratio (SNR) criterion. Numerical results show that the proposed training design considerably outperforms the baseline schemes, and the proposed passive beamforming design results in a significant improvement in performance over that of the conventional LoRa system, thereby demonstrating the feasibility of extending coverage areas of LoRa systems with the aid of IRS.

Multiobjective evolutionary algorithms are highly effective in solving multiobjective optimization problems (MOPs). The selection strategy, involving mating and environmental selection, is crucial in shaping these algorithms. However, when applied to many-objective optimization (MaOPs) with more than three objectives, existing methods face challenges due to reduced selection pressure and issues in maintaining diversity, making them less efficient. To address these challenges, we present a novel approach in this paper: the clustering-aided grid-based one-to-one selection-driven evolutionary algorithm (ClGrMOEA), designed to handle both MOPs and MaOPs effectively. The proposed ClGrMOEA introduces a hybrid approach that combines clustering-based mating selection with grid-based one-to-one environmental selection to balance convergence and diversity in MOPs and MaOPs. The mating selection employs K-means clustering to partition the objective space and utilizes a convergence indicator based on Euclidean distance to select promising solutions for offspring generation. The environmental selection combines Pareto dominance with grid-based one-to-one selection, using grid coordinate point distance to select promising solutions. An external archive based on Pareto dominance and crowding distance preserves elite individuals. Extensive experiments are conducted on 19 benchmark problems and 16 real-world problems to validate the superior performance of ClGrMOEA compared to seven state-of-the-art algorithms. The experimental results demonstrate that ClGrMOEA significantly outperforms these benchmark algorithms.

In recent decades, multi-objective evolutionary algorithms (MOEAs) have been evaluated on artificial test problems with unrealistic characteristics, leading to uncertain conclusions about their efficacy in real-world applications. To address this issue, a few benchmark test suites comprising real-world problems have been proposed for MOEAs, encompassing numerous multi-objective problems and a select few many-objective problems. Given the distinct challenges posed by many-objective optimization problems (MaOPs) and their inherent difficulty, developing a test suite that includes real-world problems with many conflicting objectives is crucial. Hence, in this paper, we propose a comprehensive test suite for benchmarking the many-objective real-world complex problems. This test suite consists of 11 problems collected from different disciplines of engineering. Furthermore, we comprehensively analyzed the problems in our newly proposed test suite, employing eight state-of-the-art algorithms rooted in various fundamental principles specifically designed to address MaOPs. The experimental findings highlight the strong performance of indicator-based, weight-vector-based decomposition, Pareto-dominance-based, and hybrid MOEAs on the proposed test suite. In contrast, reference-vector-based decomposition approaches, Pareto front shape estimation-based methods, and multi-evolution approaches exhibit relatively weaker performance.

Deep learning models learn powerful representational spaces required for handling complex tasks. Recently, data augmentation techniques, region-level, and image-level augmentation have proved effective in significantly improving deep learning models’ generalization performance. Nevertheless, such methods may lose some critical features or are still computationally heavy (or inefficient) due to additional computation burdens. To address this issue, in this paper, we present a novel unified data augmentation method for deep learning models, namely, RandMixAugment, which effectively combines the intrinsic properties of region-level augmentation and image-level augmentation. Specifically, the proposed RandMixAugment employs automated augmentation with masking and mixing operations. Experiments are conducted on well-known CIFAR datasets (CIFAR-10 and CIFAR-100) to verify the effectiveness of the proposed scheme compared to state-of-the-art augmentation techniques. The experimental results demonstrate that the proposed RandMixAugment yields superior performance over state-of-the-art techniques on image classification tasks and further improves the performance of the baseline deep learning model by 1.2% and 2.4% on CIFAR-10 and CIFAR-100 datasets, respectively.

In digital pathology, the accurate segmentation of cell nuclei in histopathology images is essential for medical image analysis. Histopathologists visually evaluate the patterns of cellular architecture and tissue patterns in histopathology image analysis for cancer detection to determine the malignant tissue portions and assess the severity of malignancy. However, manually analyzing scans using a high-resolution microscope requires significant effort and time. A computer-assisted diagnosis system utilizing deep learning (DL) algorithms rapidly, reliably, and automatically segments cell nuclei. However, the existing research studies have limited accuracy, high computational costs, and a lack of robustness and generalizability on diverse datasets. To address these issues, this paper proposes a novel and improved DL architecture based on the U-Net, namely, the CBAM-Residual U-Net for improving accuracy, robustness, and generalized segmentation algorithm that can be applied to various staining techniques and tissue structures. The proposed architecture utilizes a ResConv and convolution block attention modules (CBAM). These modules help the proposed architecture learn the image’s shallow and deep features. The CBAM module uses an attention mechanism concentrating on essential features such as cell nuclei’s shape, texture, and intensity to accurately segment the raw input patterns. The proposed CBAM-Residual U-Net involves fewer trainable parameters, reducing the computational and time cost s compared to state-of-the-art techniques. Extensive experiments and comprehensive evaluations are conducted to demonstrate the performance of the proposed scheme on publicly available datasets: i) Data Science Bowl (DSB) 2018, ii) The GlaS, iii) Triple-Negative Breast Cancer (TNBC). The experimental results show that our proposed model considerably outperforms the state-of-the-art techniques and detects cellular boundaries well, providing fine-grained segmentation results.

Many-objective optimization has recently gained popularity as it poses significant challenges for the existing algorithms. Therefore, numerous optimization algorithms have been developed to handle many-objective optimization in the literature. In addition, several studies have conducted experimental comparisons to assess the performance of optimization algorithms. Nevertheless, existing empirical studies have analyzed the performance of optimization algorithms on well-defined test problems, but it remains unclear whether the results translate to real-world scenarios. Furthermore, empirical studies on validating the performance of algorithms on real-world many-objective problems are intriguing but not yet fully explored. Therefore, in this article, we present a comprehensive comparative study evaluating the performance of 15 state-of-the-art algorithms on ten real-world many-objective applications with four to ten objectives from various domains. Further, these ten applications exhibit various mathematically challenging properties, including stochastic objectives, complex Pareto frontiers, and strong nonlinearity. In addition, four performance metrics are employed to visualize the performance of MOEAs in experimental settings. Based on comparative results, the performance of state-of-the-art algorithms with respect to different problems is evaluated herein.

In this paper, we consider a new design problem of optimizing a linear transceiver for correlated multiple-input multiple-output (MIMO) interference channels in the presence of channel state information (CSI) errors, which is a more realistic and practical scenario than those considered in the previous studies on uncorrelated MIMO interference channels. By taking CSI errors into account, the optimization problem is initially formulated to minimize the average mean square error (MSE) under the general power constraints. Since the objective function is not jointly convex in precoders and receive filters, we split the original problem into two convex subproblems, and then linear precoders and receive filters are obtained by solving two subproblems iteratively. It is shown that the proposed algorithm is guaranteed to converge to a local minimum. The numerical results show that the proposed algorithm can significantly reduce the sensitivity to CSI errors compared with the existing robust schemes in the correlated MIMO interference channel.

This paper studies the application of artificial intelligence to milling machines, focusing specifically on identifying the inputs (features) required for predicting surface roughness. Previous studies have extensively reviewed and presented useful features for surface roughness prediction. However, applying research findings to actual operational factories can be challenging due to the additional costs of sensor installations and the diverse environments present in each factory setting. To address these issues, in this paper, we introduced effective features for predicting surface roughness in situations where additional sensors are not installed in the existing environment. These features include feed per tooth, Fz; material removal rate, Q; and the load information. These features are suitable for use in highly constrained environments where separate sensor installation is not required, making it possible to apply the research findings in various factory environments. Additionally, to efficiently select the optimal subset for surface roughness prediction among subsets formed by available features, we apply causality to the feature selection method, proposing an approach called causality-driven efficient feature selection. The experimental results demonstrate that the features introduced in this paper are quite suitable for predicting surface roughness and that the proposed feature selection approach is more effective and efficient compared to existing selection methods.

A brain tumor is a disorder caused by the growth of abnormal brain cells. The survival rate of a patient affected with a tumor is difficult to determine because they are infrequent and appear in various forms. These tumors can be identified through Magnetic Resonance (MRI) Images, which plays an essential role in determining the tumor site; however, manual detection is a time-consuming and challenging procedure that can cause some errors in results. The adoption of computer-assisted approaches is essential to help in overcoming these constraints. With the advancement of artificial intelligence, deep learning (DL) models are being used in medical imaging to diagnose brain tumors using MR images. In this study, a deep convolutional neural network (CNN) EfficientNet-B0 base model is fine-tuned with our proposed layers to efficiently classify and detect brain tumor images. The image enhancement techniques are used by applying various filters to enhance the quality of the images. Data augmentation methods are applied to increase the data samples for better training of our proposed model. The results show that the proposed fine-tuned state-of-the-art EfficientNet-B0 outperforms other CNN models by achieving the highest classification accuracy, precision, recall, and area under curve values surpassing other state-of-the-art models, with an overall accuracy of 98.87% in terms of classification and detection. Other DL algorithms such as VGG16, InceptionV3, Xception, ResNet50, and InceptionResNetV2 are used for comparative analysis.

Quality control in manufacturing equipment relies heavily on the detection of steel surface defects. Recently, there have been an increasing number of efforts in which object detection techniques have been utilized to achieve promising results in the detection of steel surface defects since the defect patterns can be considered objects. To enhance the detection performance in the object detection problem, the non-maximum suppression (NMS) step, which eliminates redundant boxes overlapped with a box having the greatest detection score, is essential. In this work, we propose a novel NMS to improve the detection method of steel surface defects. The proposed NMS approach is composed of three novel techniques: IoU regularization, threshold adjustment, and comparison rule modification to enhance the detection performance. To evaluate the performance of the proposed NMS, we carry out extensive numerical experiments using the YOLOv7 and EfficientDet models on the steel surface defect datasets, NEU-DET and GC10-DET. The experimental results demonstrate that the proposed NMS outperforms the conventional NMS methods in both quantitative and qualitative manners.

Recently, deep learning has exhibited outstanding performance in various fields. Even though artificial intelligence achieves excellent performance, the amount of energy required for computations has increased with its development. Hence, the need for a new energy-efficient computer architecture has emerged, which further leads us to the neuromorphic computer. Although neuromorphic computing exhibits several advantages, such as low-power parallelism, it exhibits lower accuracy than deep learning. Therefore, the major challenge is to improve the accuracy while maintaining the neuromorphic computing-specific energy efficiency. In this paper, we propose a novel method of the inference process that considers the probability that after completing the learning process, a neuron can react to multiple target labels. Our proposed method can achieve improved accuracy while maintaining the hardware-friendly, low-power-parallel processing characteristics of a neuromorphic processor. Furthermore, this method converts the spike counts occurring in the learning process into probabilities. The inference process is conducted to implement the interaction between neurons by considering all the spikes that occur. The inferring circuit is expected to show a significant reduction in hardware cost and can afford an algorithm exhibiting a competitive computing performance.

In this paper, we study user grouping, precoding design, and power allocation for multiple-input multiple-output (MIMO) nonorthogonal multiple access (NOMA) systems. An optimization problem is formulated to the maximize the sum rate under a transmit power constraint at a base station and rate constraints on users, which are nonconvex and combinatorial and thus very challenging to solve. To tackle this problem, we carry out the optimization in two steps. In the first step, exploiting the machine learning technique, we develop an efficient iterative algorithm for user grouping and precoding design. In the second step, we develop a power-allocation scheme in closed form by recasting the original problem into a useful and tractable convex form. The numerical results demonstrate that the proposed joint scheme, including user grouping, precoding design, and power allocation, considerably outperforms the existing schemes in terms of sum rate maximization, which increases the sum-rate up to 8–18%. In addition, the results show the larger the number of antennas or users, the bigger the performance gap, at the cost of less computational complexity.

Channel estimation with received signal-to-noise ratio (SNR) feedback is promising and effective for practical wireless multiple-input multiple-output (MIMO) systems. In this paper, we investigate the channel estimation problem for the MIMO system with received SNR feedback, of which goal is to estimate the MIMO channel coefficients at a transmitter based on the received SNR feedback information from a receiver in the sense of minimizing the mean square error (MSE) of the channel estimation. For analysis, we consider two very common and widely adopted scenarios of fading: (i) quasi-static block fading and (ii) time-varying fading. In both fading scenarios, it is generally challenging to analytically tackle the channel estimation problem due to its nonlinearity and nonconvexity. To intelligently and effectively address this issue, deep learning is exploited in this paper. First, in the quasi-static block fading scenario, we propose a novel learning scheme for joint channel estimation and pilot signal design by constructing a deep autoencoder via a convolutional neural network (CNN). Also, in the time-varying fading scenario, a novel channel estimation scheme is developed by connecting a recurrent neural network (RNN) to a CNN. Moreover, in both fading scenarios, we present new and effective ways to train the proposed schemes using generative adversarial networks (GANs) to address the practical issue of a limited number of actual channel samples (i.e., real-world data) required for training. Through extensive numerical simulations, we demonstrate effectiveness and superior performance of the proposed schemes.

Coffee is the world’s most traded tropical crop, accounting for most export profits, and is a significant source of income for the countries in which it is produced. To meet the needs of the coffee market worldwide, farmers need to increase and monitor coffee production and quality. Coffee leaf disease is a significant factor that decreases coffee quality and production. In this research study, we aim to accurately classify and detect the diseases in four major types of coffee leaf disease (phoma, miner, rust, and Cercospora) in images using deep learning (DL)-based architectures, which are the most powerful artificial intelligence (AI) techniques. Specifically, we present an ensemble approach for DL models using our proposed layer. In our proposed approach, we employ transfer learning and numerous pre-trained CNN networks to extract deep characteristics from images of the coffee plant leaf. Several DL architectures then accumulate the extracted deep features. The best three models that perform well in classification are chosen and concatenated to build an ensemble architecture that is then given into classifiers to determine the outcome. Additionally, a data pre-processing and augmentation method is applied to enhance the quality and increase the data sample’s quantity to improve the training of the proposed method. According to the evaluation in this study, among all DL models, the proposed ensemble architecture outperformed other state-of-the-art neural networks by achieving 97.31% validation. An ablation study is also conducted to perform a comparative analysis of DL models in different scenarios.

In this paper, we propose a novel two-stage transformer with GhostNet, which improves the performance of the small object detection task. Specifically, based on the original Deformable Transformers for End-to-End Object Detection (deformable DETR), we chose GhostNet as the backbone to extract features, since it is better suited for an efficient feature extraction. Furthermore, at the target detection stage, we selected the 300 best bounding box results as regional proposals, which were subsequently set as primary object queries of the decoder layer. Finally, in the decoder layer, we optimized and modified the queries to increase the target accuracy. In order to validate the performance of the proposed model, we adopted a widely used COCO 2017 dataset. Extensive experiments demonstrated that the proposed scheme yielded a higher average precision (AP) score in detecting small objects than the existing deformable DETR model.

Road segmentation has been one of the leading research areas in the realm of autonomous driving cars due to the possible benefits autonomous vehicles can offer. Significant reduction of crashes, greater independence for the people with disabilities, and reduced traffic congestion on the roads are some of the vivid examples of them. Considering the importance of self-driving cars, it is vital to develop models that can accurately segment drivable regions of roads. The recent advances in the area of deep learning have presented effective methods and techniques to tackle road segmentation tasks effectively. However, the results of most of them are not satisfactory for implementing them into practice. To tackle this issue, in this paper, we propose a novel model, dubbed as TA-Unet, that is able to produce quality drivable road region segmentation maps. The proposed model incorporates a triplet attention module into the encoding stage of the U-Net network to compute attention weights through the triplet branch structure. Additionally, to overcome the class-imbalance problem, we experiment on different loss functions, and confirm that using a mixed loss function leads to a boost in performance. To validate the performance and efficiency of the proposed method, we adopt the publicly available UAS dataset, and compare its results to the framework of the dataset and also to four state-of-the-art segmentation models. Extensive experiments demonstrate that the proposed TA-Unet outperforms baseline methods both in terms of pixel accuracy and mIoU, with 98.74% and 97.41%, respectively. Finally, the proposed method yields clearer segmentation maps on different sample sets compared to other baseline methods.

In this paper, we consider an unmanned aerial vehicle (UAV) wireless communication system where a base station (BS) equipped multi antennas communicates with multiple UAVs, each equipped with a single antenna, using the LoRa (Long Range) modulation. The traditional approaches for downlink beamforming design or uplink power allocation rely on the convex optimization technique, which is prohibitive in practice or even infeasible for the UAVs with limited computing capabilities, because the corresponding convex optimization problems (such as second-order cone programming (SOCP) and linear programming (LP)) requiring a non-negligible complexity need to be re-solved many times while the UAVs move. To address this issue, we propose novel schemes for beamforming design for downlink transmission from the BS to the UAVs and power allocation for uplink transmission from the UAVs to the BS, respectively, based on deep learning. Numerical results demonstrate a constructed deep neural network (DNN) can predict the optimal value of the downlink beamforming or the uplink power allocation with low complexity and high accuracy.

Edge computing (EC) is very useful and particularly promising for many practical unmanned aerial vehicle (UAV) applications. Integrating the blockchain to this technology strengthens privacy protection and data integrity and also prevents data from being easily leaked. However, the required operations in the blockchain are computationally heavy because a blockchain requires devices to solve a complicated proof-of-work (PoW) puzzle to add new data (i.e., a block) to the blockchain. Solving a PoW requires substantial amounts of time and energy, which are big concerns for UAVs. In this article, we suggest a synergetic solution to address this issue based on multiple competing miners in a blockchain. Specifically, we present two novel frameworks for combining the blockchain and EC to effectively overcome several critical limitations when applying the blockchain to UAV and EC tasks, respectively. The goal of both of these proposed frameworks is to reduce both the time spent on mining and the energy consumption for the EC. We first look at the fundamentals of the blockchain with competing miners. Then, our proposed frameworks are described with experimental results, through which important insights are drawn. We finally discuss application scenarios for our proposed frameworks, the related technical challenges, and future research directions.

Automatic road extraction from unmanned aerial vehicle (UAV) imagery has been one of the major research topics in the area of remote sensing analysis due to its importance in a wide range of applications such as urban planning, road monitoring, intelligent transportation systems, and automatic road navigation. Thanks to the recent advances in Deep Learning (DL), the tedious manual segmentation of roads can be automated. However, the majority of these models are computationally heavy and, thus, are not suitable for UAV remote-sensing tasks with limited resources. To alleviate this bottleneck, we propose two lightweight models based on depthwise separable convolutions and ConvMixer inception block. Both models take the advantage of computational efficiency of depthwise separable convolutions and multi-scale processing of inception module and combine them in an encoder–decoder architecture of U-Net. Specifically, we substitute standard convolution layers used in U-Net for ConvMixer layers. Furthermore, in order to learn images on different scales, we apply ConvMixer layer into Inception module. Finally, we incorporate pathway networks along the skip connections to minimize the semantic gap between encoder and decoder. In order to validate the performance and effectiveness of the models, we adopt Massachusetts roads dataset. One incarnation of our models is able to beat the U-Net’s performance with 10× fewer parameters, and DeepLabV3’s performance with 12× fewer parameters in terms of mean intersection over union (mIoU) metric. For further validation, we have compared our models against four baselines in total and used additional metrics such as precision (P), recall (R), and F1 score.

Federated learning (FL) allows UAVs to collaboratively train a globally shared machine learning model while locally preserving their private data. Recently, the FL in edge-aided unmanned aerial vehicle (UAV) networks has drawn an upsurge of research interest due to a bursting increase in heterogeneous data acquired by UAVs and the need to build the global model with privacy; however, a critical issue is how to deal with the non-independent and identically distributed (non-i.i.d.) nature of heterogeneous data while ensuring the convergence of learning. To effectively address this challenging issue, this paper proposes a novel and high-performing FL scheme, namely, the hierarchical FL algorithm, for the edge-aided UAV network, which exploits the edge servers located in base stations as intermediate aggregators with employing commonly shared data. Experiment results demonstrate that the proposed hierarchical FL algorithm outperforms several baseline FL algorithms and exhibits better convergence behavior.

An experimental study on the resistance of concrete containing air-cooled slag (AS) and water-cooled slag (WS) against freeze–thaw cycles was conducted. For comparison, the durability of ASTM Type I ordinary Portland cement (OPC) concrete exposed to the same freeze–thaw environment was examined. To evaluate the durability of concrete exposed to the freeze–thaw environment, an experiment was conducted according to ASTM C 666 procedure A. Furthermore, the relative dynamic modulus of elasticity, surface electrical resistivity, and compressive strength of concrete specimens were measured after exposing them to freeze–thaw cycles for a predetermined period, and the results were compared with those of OPC concrete. The relationship between the freeze and thaw resistances of concrete and the air-void system (spacing factor and specific surface area) was identified. Furthermore, the microstructure of concrete exposed to freeze–thaw cycles was observed using scanning electron microscopy to identify the interfacial transition zone, cracks, and micropores. Experimental results showed that the resistance of blended cement concrete containing WS and AS against freeze–thaw cycles was significantly higher than that of OPC concrete. The concrete in which 10% of OPC was replaced by AS exhibited a similar durability as that of the concrete in which 40% of OPC was replaced only by WS. Therefore, it is expected that blended cement concrete containing WS and AS based on an appropriate mix proportion design will exhibit excellent durability in regions experiencing freezing temperatures.