Publications Scientifiques
Permanent URI for this communityhttps://dspace.univ-boumerdes.dz/handle/123456789/10
Browse
610 results
Search Results
Item Deformable Transformer-Based Object Detection for Robust Perception in Autonomous Driving(IEEE, 2025) Kezzal, Chahira; Benderradji, Selsabil; Benlamoudi, Azeddine; Bekhouche, Salah Eddine; Taleb, Abdel; Hadid, AbdenourAutonomous driving demands robust and real-time object detection to safely navigate in complex environments. While Convolutional neural network (CNN)-based detectors have been widely adopted, they face challenges such as limited receptive fields and inefficiencies in handling small or occluded objects. This paper presents a deformable Transformer based object detection framework designed to address these limitations. By leveraging deformable attention mechanisms, the model dynamically focuses on relevant spatial regions, significantly enhancing detection accuracy. Evaluated on the benchmark KITTI dataset, our proposed approach achieves an interesting mAP@50 of 96.6%, surpassing many state-of-the-art methods, at the cost of slower inference speed (7.0 FPS). The experimental results also demonstrate the framework’s superior precision and adaptability in autonomous driving scenarios. This work underscores the potential of deformable transformers to advance perception systems, balancing high accuracy with the demands of real-world applications.Item Efficient Real-Time Multi-Class Object Tracking with YOLO11 and ByteTrack in Real-World Driving Scenes(IEEE, 2025) Benderradji, Selsabil; Kezzal, Chahira; Benlamoudi, Azeddine; Bekhouche, Salah Eddine; Taleb, AbdelAccurate and real-time multi-object tracking (MOT) is essential for autonomous driving systems to ensure safe navigation and decision making in dynamic environments. This paper introduces a tracking-by-detection pipeline that integrates YOLOv11 a high speed, high-accuracy object detector with ByteTrack, a robust data association algorithm capable of lever-aging both high and low confidence detections. The proposed framework addresses key challenges in MOT such as frequent occlusions, fluctuating lighting, and dense traffic by combining efficient detection with motion-consistent identity tracking. Evaluated on the KITTI benchmark, our method demonstrates superior performance across multiple metrics, including HOTA, AssA, and MOTA, for both cars and pedestrians. Additionally, the system achieves an average runtime of 60.4 FPS, supporting its real-time applicability. The results confirm that the proposed YOLOv11 + ByteTrack integration provides a scalable, accurate, and deployment ready solution for complex urban driving scenarios.Item Achievable Rates of Full Duplex Cooperative Relay Selection-Based Machine Learning(IEEE, 2025) Belaoura, Widad; Althunibat, Saud; Mazen, Hasna; Qaraqe, Khalid; Ammuri, RulaMachine learning (ML) is an advanced artificial intelligence technology that addresses the ever-growing complexity in communication signal processing. In this paper, the concept of ML-based classification model to choose the best relay is investigate in a full duplex (FD) cooperative system. Specifically, a K-nearest neighbors (KNN)-based relay selection is applied to accurately predict and evaluate the achievable rate of the optimal FD relay. The core idea of the multi-class KNN is to identify the optimal relay that yields the highest achievable rate performance by utilizing a large set of offline training data derived from the channel state information (CSI), ensuring that no further training is required during system processing. The results indicate that the KNN-based FD relay selection can achieve an achievable rate comparable to the optimal exhaustive search method with lower computation complexity.Item Valuation of Physical Layer Security Under Jamming Attacks Utilizing RIS(2025) Refas, Souad; Meraihi, Yassine; Ivanova, Galina; Baiche, Karim; Cherif, Amar Ramdane; Acheli, DalilaVehicular visible light communication (V VLC) systems, when combined with reconfigurable intelligent surfaces (RIS), present promising opportunities for improving communication reliability and efficiency in vehicle to vehicle (V2V) environments. Nevertheless, safeguarding these systems at the physical layer remains a critical challenge, particularly given their exposure to jamming threats. In this study, we investigate the physical layer security performance of RIS assisted V2V VLC systems under jamming scenarios, employing realistic V2V VLC channel models. We develop a methodology to examine the impact of various security strategies in mitigating the adverse effects of jamming. Our analysis examines key parameters including the signal to noise ratio SNR, the secure communication rate and the count of RIS units. Simulation results confirm that the proposed security systems significantly enhance the resilience of V2V VLC networks in the presence of jamming attacks. These results offer useful perspectives for the reliable design structure and deployment of RIS based V2V VLC systems in practical vehicular communication settings.Item Enhancing sustainability in CNC turning of POM-C polymer using MQL with vegetable-based lubricant: machine learning and metaheuristic optimization approaches(Springer Science and Business Media, 2025) Hakmi, Tallal; Abderazek, Hammoudi; Yapan, Yusuf Furkan; Hamdi, Amine; Uysal, AlperSustainable machining of polymer parts, which is still less advanced than metal machining, aims not only to improve machinability but also to address environmental and economic challenges. Therefore, this study analyzes the sustainability of polyoxymethylene copolymer (POM-C) turning by incorporating minimum quantity lubrication (MQL) parameters (Q: flow rate, θ: nozzle angle, and d: nozzle distance) and conventional cutting parameters (Vc: cutting speed, f: feed, and ap: depth of cut), while replacing conventional oil with a biodegradable and environmentally friendly lubricant derived from Eraoil KT/2000. Additionally, the methodology relies on sustainability indicators such as surface roughness (Ra), total energy consumption (Etotal), total carbon emissions (CEtotal), and overall cost (Ctotal). To achieve this, several approaches are employed, including analysis of variance (ANOVA), artificial neural networks (ANN), k-fold cross-validation (k-fold CV), and two multi-objective metaheuristic optimization algorithms, namely SHAMODE (success history-based multi-objective adaptive differential evolution) and RPBILDE (real-code population-based incremental learning and differential evolution), are used to identify significant factors, establish mathematical models, and determine optimal conditions. The multi-objective optimization highlights trade-offs between the four sustainability criteria. Thus, a low feed value and a low MQL flow rate, combined with significant angle and distance, as well as moderate cutting speed and depth of cut, provide minimal surface roughness (Ra = 1008 µm), low energy consumption (Etotal = 0.0947 MJ), low carbon emissions (CEtotal = 0.0583 kgCO₂) but with a slightly higher cost (Ctotal = 1701 $). These results confirm a Pareto front where the improvement of one criterion negatively impacts another, guiding industrial decisions based on prioritiesItem Design and implementation of a self-driving car using deep reinforcement learning: A comprehensive study(Elsevier, 2025) Djerbi, Rachid; Rouane, Anis; Taleb, Zineb; Saradouni, SafiaThis paper presents a groundbreaking and comprehensive study on the design, implementation, and evaluation of a self-driving car utilizing deep reinforcement learning, showcasing significant advancements in autonomous vehicle technology. Our robust framework integrates three innovative AI models for essential functionalities: road detection, traffic sign recognition, and obstacle avoidance. The system architecture, structured around a three layers “DDD” (Data, Detection, Decision) approach, involves meticulous data preprocessing for traffic signs and road data, followed by specialized Deep Learning models for each detection task, including a CNN for traffic signs, a CNN for road detection, and the pre-trained MobileNet-SSD for obstacle detection. A reinforcement learning agent in the Decision Layer processes these outputs for real-time control (steering, acceleration, braking) through a continuous learning process with environmental feedback. The research encompasses both extensive simulation in Unity, leveraging the ML-Agents toolkit for agent training across diverse environments, and crucial real-world deployment. Our reward/punishment system in the simulation environment, based on collisions with road markers and obstacles, refined the agent's decision-making. The trained AI models were successfully exported and deployed onto a physical prototype, controlled by a Raspberry Pi and equipped with a camera and ultrasonic sensors. Real-world testing affirmed the robust performance of the physical model in detecting roads, recognizing traffic signs, and effectively avoiding obstacles. Quantitative results demonstrate compelling performance, including over 90% accuracy in obstacle detection and a 15% improvement in navigation efficiency compared to traditional algorithms under controlled simulation conditions. Model evaluation metrics show a 98% accuracy, 12% loss, and a prediction rate exceeding 77%. This study not only contributes a comprehensive framework for autonomous vehicle development but also highlights the transformative potential of deep reinforcement learning for creating intelligent and adaptable autonomous systems in both virtual and real-world scenarios, paving the way for safer and more efficient transportation technologiesItem Rigorous Explainable Artificial Intelligence Models for Predicting CO2-Brine Interfacial Tension: Implications for CO2 Sequestration in Saline Aquifers(American Chemical Society, 2025) Nait Amar, Menad; Youcefi, Mohamed Riad; Alqahtani, Fahd Mohamad; Djema, Hakim; Ghasemi, MohammadCarbon capture and sequestration (CCS) is an attractive approach for reducing carbon dioxide (CO2) emissions, with saline aquifers offering promising sites for long-term sequestration. Interfacial tension (IFT) between CO2 and brine plays a crucial role in the trapping efficiency. This study develops explainable artificial intelligence (XAI) models to accurately predict the IFT in CO2–brine systems. Three advanced machine learning models, namely, Super Learner (SL), Elman Neural Network (ENN), and Power Law Ensemble Model, were implemented based on a data set comprising 2616 measurements. Among the established paradigms, SL achieved the highest accuracy (RMSE = 0.7813 and R2 = 0.9953) across diverse conditions. To enhance model transparency, Local Interpretable Model-agnostic Explanations and SHAP (SHapley Additive Explanations) interpretability techniques were employed, confirming strong alignment with experimental trends. Comparative analysis further demonstrated that the SL scheme surpasses existing literature models. Overall, this study highlights the effectiveness of XAI-based predictive modeling for accurately estimating the CO2–brine IFT under diverse operational conditions. Future implementation in real CCS projects can offer valuable insights into injection strategies, trapping mechanisms, and long-term formation stabilityItem Enhancing Data Privacy in Intrusion Detection: A Federated Learning Framework With Differential Privacy(John Wiley and Son, 2025) Saidi, Ahmed; Khouri, A. OuadoudThe rise of cyber threats has underscored the critical need for robust intrusion detection systems (IDS). While traditional approaches, including statistical, knowledge-based, and AI-driven methods, have been pivotal, they often face limitations such as data privacy concerns, scalability challenges, and low detection accuracy on unfamiliar threats. This paper addresses these issues by adopting a federated learning (FL) paradigm for collaborative intrusion detection, allowing data to remain local and enhancing privacy protection. The proposed solution integrates advanced encryption techniques and differential privacy to safeguard confidentiality while ensuring system scalability and adaptability. By introducing a robust separation of agents' roles and leveraging FL's decentralized architecture, the system overcomes the limitations of centralized learning, including single points of failure and communication overhead. Experimental results validate the proposed architecture, demonstrating significant improvements in performance and offering a promising direction for modern network security. This work not only highlights the potential of FL-based IDS but also explores the integration of distributed ledger technologies to further enhance trust and security. These findings contribute to the growing field of privacy-preserving computing and lay the groundwork for future innovations in scalable, secure, and efficient intrusion detection systemsItem Androgen receptor expression in triple negative breast cancer: an Algerian population study(Taylor and Francis, 2025) Hedjem, Amel; Kouchkar, Amal; Ladjeroud, Amel; Zerrouki, Nacera; Benaissa, Fatima; Ibrahim, Nasir A.; Aleissa, Mohammed Saad; Basher, Nosiba S.; Derguini, Assia; Idres, TakfarinasTriple-negative breast cancer (TNBC) is a molecular subtype of breast cancer characterized by the absence of estrogen and progesterone receptors and the lack of HER2 overexpression. TNBC is highly heterogeneous, complicating the identification of new therapeutic targets. However, the expression of the androgen receptor (AR) in the luminal androgen receptor (LAR TNBC) subgroup has opened the door to alternative therapeutic approaches. This study aimed to assess AR expression and correlate it with clinicopathological factors in 160 early-stage TNBC patients treated from February 2015 to February 2017. Our findings reveal that AR expression is observed in 16.87% (27/160) of ≥1% AR positivity cases. Moreover, a significant 12.5% (20/160) was found in ≥10% AR positive cases. Positive AR expression was inversely correlated with a high Ki-67 proliferation index and with the basal immunophenotype. The five-year survival rate for our cohort was 83.12%, and no significant association between AR expression and overall survival was observed (p = 0.77). The study highlights the potential role of AR expression in TNBC and its implications for therapeutic strategies, although no significant association with overall survival was foundItem An improved artificial neural network using weighted mean of vectors algorithm for precise GTAW weld quality prediction and parameter optimization(Springer Science and Business Media, 2026) Boucetta, Brahim; Boumediene, Faiza; Ait Chikh, Mohamed Abdessamed; Afia, AdelAccurate prediction of mechanical properties in gas tungsten arc welding (GTAW) remains challenging due to the complex, nonlinear relationships between process parameters and weld quality. This study introduces a novel framework that systematically evaluates seven state-of-the-art metaheuristic algorithms: spider wasp optimizer (SWO), weighted mean of vectors (INFO), gradient-based optimizer (GBO), artificial rabbits optimization (ARO), blood-sucking leech optimizer (BSLO), RUN beyond the metaphor (RUN), and successive history adaptive differential evolution (SHADE), for training artificial neural networks (ANNs) to predict ultimate tensile strength in GTAW of Inconel 825 alloy. The primary novelty lies in identifying the gradient-based optimizer as the most effective algorithm for this application, presenting superior generalization capability and establishing a new benchmark for welding parameter prediction. The optimized ANN-GBO model achieved significant performance improvements over conventional ANN approaches, with the coefficient of determination () increasing from 0.6844 to 0.8669 (26.7% improvement) and root mean square error (RMSE) decreasing from 51.89 MPa to 33.71 MPa (35.0% reduction). These substantial enhancements in prediction accuracy provide critical insights for optimizing high-performance nickel-based alloy welding processes