Volume 35, Issue 1

Mobile robot navigation remains a critical challenge in robotics, with applications spanning autonomous vehicles, search and rescue, and other dynamic environments. In recent years, reinforcement learning (RL) has become a powerful approach for solving complex tasks such as robotic manipulation, gameplay, and autonomous driving. By enabling robots to learn optimal navigation strategies through interaction with their environment, RL offers a promising pathway to autonomous mobility. This review presents a comprehensive overview of recent advancements in RL as applied to mobile robot navigation. We begin by outlining core RL concepts, agents, environments, rewards, and value functions, explaining their roles in navigation. Key RL techniques, including Q-learning, deep reinforcement learning, Markov Decision Processes (MDPs), and policy gradient methods, are examined to highlight their transformative impact on navigation performance. The review also explores a range of practical applications and identifies current challenges and open research directions. Critical issues such as safety, sample efficiency, and scalability to real-world scenarios are discussed in depth to ensure robust deployment of RL-based systems. Lastly, this review synthesizes the state-of-the-art in reinforcement learning for mobile robot navigation, offering readers both a foundational understanding and valuable insights into emerging trends and future opportunities in this rapidly evolving field.

Early identification of risks improves patient outcomes and overall health. The early identification of high-risk status helps in early medical attention and prevents complications. Understanding a patient’s risk level enables better management of chronic conditions. Identifying high-risk patients is essential for the efficient allocation of healthcare resources. This approach focuses on predicting abnormalities by analysing the correlation between multiple vital signs. This review highlights body temperature, heart rate, respiratory rate, oxygen saturation, and blood pressure as important vital signs for determining a patient’s risk level. A total of 180 articles were reviewed, of which 20 were selected for inclusion. This study focuses mainly on research conducted from 2016 to 2024. The identified studies focused on critical illnesses (4, 20 %), cardiovascular diseases (4, 20 %), chronic conditions (3, 15 %), abnormalities (3, 15 %), Coronavirus Disease 2019 (COVID-19) (2, 10 %), mortality (2, 10 %), falls (1, 5 %), and hypertension (1, 5 %). Most studies have focused on physiological parameters, such as oxygen saturation, blood pressure, heart rate, respiratory rate, and body temperature. The severity level of a patient can be determined and treated by monitoring these vital signs. Developments in wearable sensor technology have facilitated the continuous monitoring of vital signs and assisted healthcare professionals.

A rapid rise in machine learning-based applications has made it one of the most popular areas in the field of artificial intelligence (AI). The most commonly used libraries to implement the algorithms used in these applications are Scikit learn and Weka. It is challenging to test these machines learning based applications due to the Oracle Problem. The problem is when the expected outcome is not known and hence the testing of such applications cannot be performed via traditional testing techniques. One of the solution to the Oracle problem is the use of Metamorphic testing to test the machine learning applications. The code of machine learning algorithms is often ignored, when testing of ML-based applications is done. However, the usage of the machine learning algorithms within the libraries requires formal testing to improve reliability. This work evaluates the Metamorphic relations for machine learning algorithms by finding their kill rate while testing 5 machine learning (ANN, ID3, KNN, Naive Bayes, SVM) classifiers from the Scikit Learn library. This work also calculates the statement coverage, while testing the metamorphic relations. The relationship between the effectiveness of fault detection and code coverage is identified as well.

Asphalt roads play a vital role in land transportation systems, significantly contributing to the growth and development of economies and societies. However, over time, the quality of these roads deteriorates due to aging and the cumulative effects of wear and tear, leading to various pavement and road issues such as potholes, cracks, and damaged sidewalks. This paper aims to develop a deep learning model, specifically leveraging the YOLOv8 object detection framework, to detect and classify road infrastructure problems using images captured from unmanned aerial vehicles (UAVs). The model processes a series of road images from the Roboflow dataset and was trained on Google Colab, utilizing advanced machine learning techniques to analyze the images and accurately identify road damage. Subsequently, the model was evaluated using metrics such as accuracy, recall, precision, and F1-score. The results demonstrated that the model is both efficient and reliable. The model achieved high performance, with an F1-score of 94 %, precision of 93 %, and recall of 95 %, which indicates its effectiveness in identifying various road defects. By detecting and locating issues such as potholes, cracks, and sidewalk damage, this model offers a promising solution for maintaining road infrastructure, supporting smart transportation systems, enhancing road safety, and helping reduce hazards and accidents.

This paper proposes a computerized text analysis framework to examine the evolution of China’s industrial internet policies concerning SMEs. Analyzing a corpus of 310 policy documents concerning SMEs in industrial internet domain, we first investigated their external structure. Thematic sequence evolution analysis subsequently identified five key topics. Building on a pivotal policy issued in 2017, the evolution of China’s industrial internet policies concerning SMEs was categorized into two stages. The paper further explored the characteristics of the intergovernmental cooperation network, and examined the evolutionary paths of policy themes using a Sankey diagram. The findings indicate that the five topics in stage 2 are essentially the continuation and further development of the policy themes in stage 1, demonstrating strong policy coherence in industrial internet domain. Theoretically, this study addresses the IAD’s challenges in exploring the evolution of action situations and the diverse processes of structural changes over time, while overcoming the ACF’s limitations in interpreting the interactions of common belief systems and interests.

Traffic sign recognition plays a critical role in the development of autonomous driving systems and intelligent transportation networks. However, detecting small traffic signs in real-world scenarios, particularly those captured from vehicle-mounted cameras, remains a significant challenge due to their diminutive size, low resolution, and environmental noise. To address these challenges, we propose an innovative multi-strategy enhancement framework for YOLOv7, designed specifically to improve small target detection. The framework integrates several novel techniques: the SE attention mechanism is incorporated into the ELAN module of the backbone network to enhance feature discriminability; DySample replaces traditional upsampling methods in the head network to refine feature reconstruction; NWD loss is introduced as a superior alternative to CIoU loss, improving the localization accuracy of small objects; and PConv convolution is applied to reduce model parameters without sacrificing performance. Experimental results on the CCTSDB-2021 dataset demonstrate the effectiveness of these enhancements, with mAP@0.5 and mAP@0.5:0.95 improving by 6.6  and 15.2 %, respectively, compared to the original YOLOv7 model. The proposed algorithm outperforms YOLOv7 by 10.9 % in mAP@0.5 and by 11.89 % in mAP@0.5:0.95 on the TT100K dataset. Moreover, the optimized model achieves real-time inference at 83 FPS on the CCTSDB-2021 dataset, while reducing the number of parameters by 1.5 million, making it highly efficient for practical deployment in autonomous vehicles. These improvements not only enhance detection accuracy but also meet the real-time processing requirements of intelligent transportation systems.

Deep learning techniques have been widely used in various fields. However, they face significant security challenges due to the existence of adversarial examples. Traditional black-box adversarial attack methods mainly rely on swarm intelligence optimization algorithms to identify optimal perturbation pixels, which requires intensive computational resources. In some typical applications such as medical image recognition, public datasets are often used to train deep learning models. It is worth noting that such dataset inherently contains some basis features for deep learning models to learn discriminative representations. And these features can serve as critical cues for constructing adversarial samples. Inspired by this observation, a novel adversarial attack method was proposed. First, some sensitive locations are identified within the dataset without querying the target model. Moreover, the adversarial attack samples are constructed based on these locations. Different from white-box and black-box attack, dataset characteristics are utilized to construct adversarial attack samples. The proposed method investigates naturally occurring vulnerabilities in the data, offering new insights for enhancing data augmentation techniques and attack strategies, while also providing a promising direction for improving model robustness. Experimental results demonstrate that this method can achieve attack effectiveness comparable to the Particle Swarm Optimization (PSO) algorithm.

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that leads to a decline in memory and cognitive functions. An early and accurate diagnosis is critical for effective management and treatment also tend to lack sufficient accuracy. Accurate and early discrimination of Alzheimer’s disease (AD) remains a critical challenge in medical imaging and computational neuroscience. Traditional diagnostic approaches, such as clinical assessments and neuroimaging, are often subjective and labor-intensive work. Recently, convolutional neural networks (CNNs) and handcrafted feature extraction techniques have shown promising results for automated AD classification. In order to improve AD detection accuracy, this study suggests a novel and efficient hybrid feature extraction method to address the accurate detection of Alzheimer’s disease (AD). This method combines deep feature representations taken from customized CNN with the Generalized Hadamard Difference (GHD) operator, a mathematical technique that is intended to capture subtle structural variations. This integrated approach leverages the complementary strengths of handcrafted and learned features to better characterize the complex patterns associated with AD progression. The publicly accessible OASIS-1 MRI dataset was employed to rigorously evaluate the proposed method, consisting of high-resolution T1-weighted brain images from cognitively normal and impaired subjects. Classification was performed using a Support Vector Machine (SVM) classifier, yielding an impressive overall accuracy of 99.50 %, surpassing many existing state-of-the-art methods. The results highlight that incorporating GHD with deep features improved the method’s ability to detect early and subtle manifestations of AD.

This study aims to develop an automatic system for detecting toxic comments in online environments, particularly on social networking platforms. The focus is on efficiently identifying and categorizing toxicity in user-generated comments to address the growing issue of harmful online content. A novel hybrid model, Text-BGRU-CNN, combining Bidirectional Gated Recurrent Unit (BGRU) and Text Convolutional Neural Network (Text-CNN), is introduced for multilabel toxicity detection. This model uses pre-trained word embeddings for word vector generation and a range of filters to extract local features and long-term dependencies in text. It incorporates a fully connected layer, a normalization layer, and an output layer for multilabel category prediction. The proposed hybrid model demonstrates superior classification accuracy in experimental trials. It was tested on a dataset divided into training and testing sets, enhanced by significant pre-processing. Structural modifications, including increasing dense units and filters, were evaluated. The final model, combining GRUs with a single CNN layer, achieved an accuracy of 0.9944 in classifying toxic comments. The study evidences the efficacy of a hybrid GRU and single-layer CNN model in online toxic comment classification. Results suggest that simpler model architectures, supplemented by extensive pre-processing, yield high accuracy and efficient training. The findings underscore the importance of further research to understand biases in trained classifiers and suggest exploring alternative methods for representing sequences and independent training of sparsely labeled classes in future work.

Cross-project defect prediction is an important method of identifying defects in a project. We extract knowledge from the source project and apply it to predict labels for the target project in cross-project defect prediction. However, insignificant and irrelevant features can have an impact on model performance. By selecting only significant and relevant features, hybrid feature selection can help achieve high prediction accuracy. Our goal is to investigate the effect of significant feature selection along with CNN classifier on cross-project defect prediction for multi-class datasets using a hybrid approach. We took advantage of the strengths of Random Forest and Recursive Feature Elimination Cross Validation, which can constructively select a few significant features. Our controlled experiment has a 1 Factor 2 Treatments design. Exploratory Data Analysis demonstrates that all versions of the PROMISE repository are multi-class and contain duplicated rows of data, a distribution gap between values, and imbalance classes. After removing duplicated rows, narrowing the data distribution gap, and balancing classes, we chose a significant feature set using a hybrid approach that included Random Forest and Recursive Feature Elimination Cross Validation. To predict Cross project defects, we used Convolutional Neural Network as a classifier, with SoftMax as the final layer. In terms of Area Under Curve, our experimental setup resulted in an average 78 % prediction accuracy measure across all 14 versions. Our experimental results showed that CNN classifier using features selected through Hybrid Feature Selection has a significant impact on defect prediction accuracy for different datasets in the Cross Project.

The binary classification of three-dimensional (3-D) objects for phase-only digital holographic information is performed using the various deep learning network models such as ResNet50, ResNet101, ResNet152, ResNet18, ResNet34, EfficientNetB0, DenseNet121, DenseNet169, Neural Architectural Search (NAS) Network, and InceptionV3. The four 3-D objects considered to perform binary classification are ‘triangle-square’, ‘circle-square’, ‘square-triangle’, and ‘triangle-circle’. The 3-D object ‘triangle-square’ has been considered for Class 1 and the remaining three 3-D objects have been considered for Class 2. The digital holograms of 3-D objects have been formed using the phase-shifting digital holographic (PSDH) technique and numerically reconstructed to obtain phase images. The phase image dataset consisting of 2,880 images was trained using all the various deep learning network models to obtain the results. The results such as loss/accuracy, loss/positive predictive value (PPV), and loss/sensitivity curves on the training/validation sets, error matrix, and performance metrics namely accuracy, PPV, sensitivity, F1-score, Matthews correlation coefficient (MCC), cohen_kappa_score (CKS), balanced_accuracy_score (BAS), jaccard_score (JS), log_loss (LL), hinge_loss (HL), and brier_score_loss (BSL) are shown for the binary classification task. Finally, the results such as receiver operating characteristic (ROC), and PPV-sensitivity curve are also shown to justify the performance of the work. The results obtained from the deep residual network models i.e. ResNet50, ResNet101, ResNet152, ResNet18, and ResNet34 were compared with other deep learning network models such as EfficientNetB0, DenseNet121, DenseNet169, NAS Network, and InceptionV3.

This study addresses the diagnostic challenges associated with Non-Small Cell Lung Cancer (NSCLC), the most prevalent form of lung cancer often diagnosed at advanced stages. It aims to develop a computer-aided classification model exclusively utilizing medical images from Positron Emission Tomography (PET) and Computed Tomography (CT) scans. The model identifies benign/malignant Solitary Pulmonary Nodules (SPN) related to NSCLC. A dataset comprising of 456 patients, in total, was curated, featuring 48.68 % benign cases. To achieve its objective, four well-established Deep Learning (DL) algorithms were employed. The dataset was split into three different groups of images, each used for a particular task; training, testing and validation of the model. Notably, the study extends beyond predictive accuracy by delving into the prediction process of the best-performing model, thereby enhancing the explainability of the typically opaque Artificial Intelligence (AI) models. This explainability aspect aims to foster trust and confidence in the model’s outcomes, allowing users to comprehend the decision-making process. The results indicate that the YOLOv8 algorithm emerged as the most accurate classification model, achieving a maximum accuracy of 91.3 % and a maximum True Positive Rate (TPR) of 93.62 %. This study’s importance lies in underscoring the potential of DL approaches in improving NSCLC diagnosis while providing a transparent and understandable classification mechanism. It offers a novel way of explaining classification results from YOLOv8 model and it demonstrates both the effectiveness of DL-assisted predictions in characterizing SPNs and the added value of interpretability, thereby offering a holistic perspective on the model’s capabilities.

Accurate classification of child delivery mode is crucial for improving maternal and neonatal health. In developing countries like Ethiopia, clinical assessments alone often result in misguided medical interventions. Even though machine learning in healthcare has brought promises, the various algorithms along with different real-world datasets perform differently. Hence, the objective of this study was to develop a machine learning model for predicting child delivery mode based on real data. The study followed experimental and exploratory research design utilizing 1,072 antenatal records from Arba Minch General Hospital and Birbir Health Center, Ethiopia. 16 attributes were considered including the outcome, mode of delivery. Predictors included sociodemographic and clinical variables such as age, weight, blood pressure, previous CS, and fetal presentation. Five machine learning algorithms including Logistic Regression, Support Vector Machine, Random Forest, Gradient Boosting, and CatBoost were trained and evaluated using hold-out validation. Additionally, a recent deep learning model, TabPFN, and Long Short-Term Memory were examined to expand the exploration. The results showed that RF (93.1 %) achieved the highest accuracy. TabPFN scored the second best accuracy score (92.5 %), demonstrating its potential on smaller tabular data. LSTM performed better than SVM and LR highlighting the consideration of the inherent temporal characteristics of the data.

The need for efficient logistics management and supply chain optimization has grown with the rapid expansion of global trade and e-commerce. Traditional logistics methods often face inefficiencies, high costs, and slow response times, making it necessary to explore innovative solutions. This study integrates artificial intelligence (AI) and Internet of Things (IoT) technologies to optimize logistics processes in transportation and inventory management. By using the random forest algorithm, this research focuses on improving transportation route planning, reducing inventory costs, and enhancing order processing efficiency. Real-time sensor data from IoT devices such as GPS and RFID tags provided valuable inputs, ensuring accurate and up-to-date information for decision-making. The results demonstrate a 28 % reduction in transportation time, a 16.7 % decrease in fuel consumption, and a 20.2 % reduction in inventory holding costs. Additionally, customer satisfaction increased by 14.1 % points. These findings indicate that the integration of AI and IoT offers significant improvements in logistics management, presenting a promising avenue for cost reduction and enhanced customer service.

Deep Vein Thrombosis (DVT) is a serious health issue in which a thrombus (blood clot) forms in one of the veins (typically in the legs). This condition can lead to various severe complications, such as post-thrombotic syndrome, pulmonary embolism, chronic vein insufficiency, and vein gangrene. It is challenging to diagnose DVT in advance since it can also occur without causing any symptoms. Hence, this study uses explainable machine learning techniques to predict DVT. Explainable Artificial Intelligence (XAI) techniques make the algorithms more robust due to their interpretability and transparency. Five different explainers have been utilized in this research, and according to them, ankle swelling, leg swelling, immobilization, and edema are the most essential parameters to identify DVT in patients. The explainable models can be used to predict DVT in advance so that appropriate care and medical attention can be provided to prevent the severe complications induced by this deadly condition.

In the realm of industrial automation, innovative AI-driven solutions are revolutionizing object detection and counting processes. This study presents a modified YOLOv10 model enhanced with the Ghost mechanism, including GhostConv and C3Ghost modules, designed to optimize computational efficiency while achieving superior detection accuracy. The proposed model excels in real-time applications, delivering a precision of 0.972, recall of 0.967, and mean Average Precision (mAP) scores of mAP@50=0.991 and mAP@50-95=0.799, all while reducing the parameter count to 6.5 M. These advancements address the challenges of fastener detection, particularly in cluttered environments and under diverse lighting conditions, paving the way for streamlined operations in manufacturing assembly lines. By leveraging specialized datasets tailored to factory-specific conditions and incorporating advanced algorithmic improvements, the model demonstrates its capacity to enhance inventory management and quality control processes. This study underscores the importance of lightweight yet robust AI models in modern manufacturing, setting a benchmark for scalable and efficient automation systems that cater to diverse industrial needs.