Saleh Yousefi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Saleh Yousefi has a total of 62 co-authors. Here are the top co-authors by number of shared publications:
- Masoud Asghari
Department of Computer Engineering, Faculty of Engineering, University of Maragheh, P.O. Box 55136-553, Maragheh, Iran
6 shared publications - Thomas Kunz
Carleton University Ottawa Canada
5 shared publications Dusit Niyato
Nanyang Technological University (NTU)
3 shared publicationsJamshid Bagherzadeh
3 shared publications- Abderrahim Benslimane
LIA Laboratory, Avignon University, Avignon, France
2 shared publications - Asghar Asgharian Sardroud
Urmia University
2 shared publications
Latest Publications
- Open Access
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
Age of Information-Aware Data Collection for Multi IRS-Assisted UAV Networks: A Deep Reinforcement Learning Approach
The use of unmanned aerial vehicles (UAVs) for collecting data from sensors has garnered considerable attention in both academic and industrial fields in recent years. To assess the freshness of the information gathered, the age of information (AoI) metric has emerged as an important tool, particularly in time-sensitive applications such as traffic management or monitoring critical issues like fire hazards. This study explores a scenario where a UAV, supported by multiple intelligent reflecting surfaces (IRSs), periodically gathers data from distributed internet of things devices (IoTDs). To minimize the average AoI across all devices, a comprehensive optimization approach is employed. Key elements such as the UAV’s trajectory, device transmission scheduling, timeslot duration, IRS selection factor, and the phase shift matrix of the selected IRS are jointly optimized. However, the problem’s complexity arises from its mixed-integer nature—IRS selection and device scheduling coefficients require integer values—and the strong interdependence of channel conditions with the UAV’s flight path and IRS phase shifts. These challenges make the problem difficult to solve. To address this, the primary optimization problem is framed as a Markov decision process (MDP). A deep reinforcement learning (DRL)-based approach is then adopted to determine the optimal UAV trajectory, device scheduling, and timeslot duration. Additionally, a location-based method is employed for selecting the suitable IRS, while a closed-form solution is used to design the phase shift matrix of the selected IRS. Results evidence improvements in successful transmission and fairness selection of devices when using multiple IRSs and variable timeslots.
- 12 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
Open AccessAge of Information-Aware Data Collection for Multi IRS-Assisted UAV Networks: A Deep Reinforcement Learning Approach
The use of unmanned aerial vehicles (UAVs) for collecting data from sensors has garnered considerable attention in both academic and industrial fields in recent years. To assess the freshness of the information gathered, the age of information (AoI) metric has emerged as an important tool, particularly in time-sensitive applications such as traffic management or monitoring critical issues like fire hazards. This study explores a scenario where a UAV, supported by multiple intelligent reflecting surfaces (IRSs), periodically gathers data from distributed internet of things devices (IoTDs). To minimize the average AoI across all devices, a comprehensive optimization approach is employed. Key elements such as the UAV’s trajectory, device transmission scheduling, timeslot duration, IRS selection factor, and the phase shift matrix of the selected IRS are jointly optimized. However, the problem’s complexity arises from its mixed-integer nature—IRS selection and device scheduling coefficients require integer values—and the strong interdependence of channel conditions with the UAV’s flight path and IRS phase shifts. These challenges make the problem difficult to solve. To address this, the primary optimization problem is framed as a Markov decision process (MDP). A deep reinforcement learning (DRL)-based approach is then adopted to determine the optimal UAV trajectory, device scheduling, and timeslot duration. Additionally, a location-based method is employed for selecting the suitable IRS, while a closed-form solution is used to design the phase shift matrix of the selected IRS. Results evidence improvements in successful transmission and fairness selection of devices when using multiple IRSs and variable timeslots.
- 25 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (4) , 5869-5884
Optimized UAV Deployment and Blockchain-Based Caching: A Reinforcement Learning Framework
Within the rapidly expanding Internet of Vehicles (IoV) landscape, the demand for network services to accommodate data-intensive applications has become increasingly paramount. However, IoV faces significant challenges, including high latency, limited coverage in remote areas, network congestion, and privacy concerns in content popularity prediction. To address these challenges, we introduce the scalable optimization for networked aerial vehicles (SONA) scheme, which reduces latency through advanced caching techniques while enhancing coverage and ensuring required data rates using uncrewed aerial vehicles (UAVs). UAVs are deployed to augment coverage in areas lacking roadside unit (RSU) support and to assist in scenarios where RSUs are overwhelmed, ensuring continuous data rate provision. Our scheme introduces a novel mathematical optimization model and machine learning algorithms: federated learning (FL) to collaboratively predict content popularity without exposing user data, reinforcement learning (RL) to dynamically optimize UAV placement and energy-efficient deployment, and blockchain to enable secure, decentralized coordination between RSUs and UAVs for real-time decision-making. Simulation results demonstrate the effectiveness of our approach, achieving an average delay of 8 ms, an average cache hit rate of 88.43%, and satisfying desired data rate requirements in 84.85% of scenarios, significantly improving IoV performance.
- 25 November 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (4) , 5869-5884
Optimized UAV Deployment and Blockchain-Based Caching: A Reinforcement Learning Framework
Within the rapidly expanding Internet of Vehicles (IoV) landscape, the demand for network services to accommodate data-intensive applications has become increasingly paramount. However, IoV faces significant challenges, including high latency, limited coverage in remote areas, network congestion, and privacy concerns in content popularity prediction. To address these challenges, we introduce the scalable optimization for networked aerial vehicles (SONA) scheme, which reduces latency through advanced caching techniques while enhancing coverage and ensuring required data rates using uncrewed aerial vehicles (UAVs). UAVs are deployed to augment coverage in areas lacking roadside unit (RSU) support and to assist in scenarios where RSUs are overwhelmed, ensuring continuous data rate provision. Our scheme introduces a novel mathematical optimization model and machine learning algorithms: federated learning (FL) to collaboratively predict content popularity without exposing user data, reinforcement learning (RL) to dynamically optimize UAV placement and energy-efficient deployment, and blockchain to enable secure, decentralized coordination between RSUs and UAVs for real-time decision-making. Simulation results demonstrate the effectiveness of our approach, achieving an average delay of 8 ms, an average cache hit rate of 88.43%, and satisfying desired data rate requirements in 84.85% of scenarios, significantly improving IoV performance.
- Open Access
- 24 November 2025
- Published by Wiley in Health Science Reports
- Vol. 8 (11)
Improving Sepsis Mortality Prediction With Machine Learning Using Full Region Synthetic Sampling Approach
Background and Aims: Sepsis is a life‐threatening condition and remains a leading cause of mortality in Intensive Care Units (ICUs). Accurate mortality prediction is crucial for optimizing ICU resource allocation. However, severe class imbalance in ICU datasets hampers the generalization performance of machine learning models. Methods: This study proposes a Full Region Synthetic Sampling Approach (FRSSA), a novel data augmentation method that dynamically balances the minority class distribution based on regional density. Additionally, we introduce Adaptive Synthetic Sampling Tuning (ASST), an optimization‐based strategy that adjusts augmentation weights to enhance model performance. To evaluate model fairness and clinical utility, we propose the Balanced Performance Score (BPS), which integrates accuracy, precision, and recall for personalized ICU risk assessment. Also, we compare two augmentation strategies: (1) Pre‐Splitting: Augmentation occurs before data set splitting, and (2) Post‐Splitting: Augmentation is applied only to the training set to ensure fair evaluation. We utilize publicly available ICU datasets from both the MIMIC‐IV and eICU‐CRD databases and evaluate the performance of Random Forest, XGBoost, and LightGBM models, with hyperparameters optimized using RandomizedSearchCV on the training set. Results: The Pre‐Splitting strategy achieved 89.64% accuracy and an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.968, but posed a risk of test‐set contamination. In contrast, the Post‐Splitting strategy yielded 78.31% accuracy and an AUROC of 0.7241, ensuring better real‐world generalization. The ASST mechanism optimally balanced 65% interpolation (FRSSA) and 35% expansion (FRSSA), reducing false positive rates and enhancing model fairness. Conclusion: FRSSA preserves regional data distribution by generating synthetic samples near the imbalanced regions. ASST dynamically adjusts augmentation ratios to maximize classification performance and generalization. The BPS score‐fusion metric offers a flexible evaluation framework, accommodating varying clinical priorities in ICU settings. Our findings demonstrate that Post‐Splitting augmentation with FRSSA and ASST produces a fairer and more reliable ICU mortality prediction model.
- 24 November 2025
- Published by Wiley in Health Science Reports
- Vol. 8 (11)
Open AccessImproving Sepsis Mortality Prediction With Machine Learning Using Full Region Synthetic Sampling Approach
Background and Aims: Sepsis is a life‐threatening condition and remains a leading cause of mortality in Intensive Care Units (ICUs). Accurate mortality prediction is crucial for optimizing ICU resource allocation. However, severe class imbalance in ICU datasets hampers the generalization performance of machine learning models. Methods: This study proposes a Full Region Synthetic Sampling Approach (FRSSA), a novel data augmentation method that dynamically balances the minority class distribution based on regional density. Additionally, we introduce Adaptive Synthetic Sampling Tuning (ASST), an optimization‐based strategy that adjusts augmentation weights to enhance model performance. To evaluate model fairness and clinical utility, we propose the Balanced Performance Score (BPS), which integrates accuracy, precision, and recall for personalized ICU risk assessment. Also, we compare two augmentation strategies: (1) Pre‐Splitting: Augmentation occurs before data set splitting, and (2) Post‐Splitting: Augmentation is applied only to the training set to ensure fair evaluation. We utilize publicly available ICU datasets from both the MIMIC‐IV and eICU‐CRD databases and evaluate the performance of Random Forest, XGBoost, and LightGBM models, with hyperparameters optimized using RandomizedSearchCV on the training set. Results: The Pre‐Splitting strategy achieved 89.64% accuracy and an Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.968, but posed a risk of test‐set contamination. In contrast, the Post‐Splitting strategy yielded 78.31% accuracy and an AUROC of 0.7241, ensuring better real‐world generalization. The ASST mechanism optimally balanced 65% interpolation (FRSSA) and 35% expansion (FRSSA), reducing false positive rates and enhancing model fairness. Conclusion: FRSSA preserves regional data distribution by generating synthetic samples near the imbalanced regions. ASST dynamically adjusts augmentation ratios to maximize classification performance and generalization. The BPS score‐fusion metric offers a flexible evaluation framework, accommodating varying clinical priorities in ICU settings. Our findings demonstrate that Post‐Splitting augmentation with FRSSA and ASST produces a fairer and more reliable ICU mortality prediction model.
- 28 October 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (16) , 1-73
Slicechain: a blockchain-based multi-domain resource orchestration in core network slicing
- 28 October 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (16) , 1-73
Slicechain: a blockchain-based multi-domain resource orchestration in core network slicing
- 15 April 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (5) , 1-18
DDos prevention in IoT networks by analyzing source-side inter-bot traffic using deep learning techniques
- 15 April 2025
- Published by Springer Nature in The Journal of Supercomputing
- Vol. 81 (5) , 1-18
DDos prevention in IoT networks by analyzing source-side inter-bot traffic using deep learning techniques
- Open Access
- 4 January 2025
- Published by Association for Computing Machinery (ACM)
- p. 11-20
Multi-Stream TSN Gate Control Scheduling in the Presence of Clock Synchronization
- 4 January 2025
- Published by Association for Computing Machinery (ACM)
- p. 11-20
Open AccessMulti-Stream TSN Gate Control Scheduling in the Presence of Clock Synchronization
- Open Access
- 27 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Networking Letters
- Vol. 7 (1) , 41-45
Latency Bounds for TSN Scheduling in the Presence of Clock Synchronization
The IEEE 802.1Qbv (80.21Qbv) standard is designed for traffic requiring deterministic and bounded latencies through strict periodic time synchronization, as specified by IEEE 802.1AS standard. However, internal clock drift in devices causes timing misalignment, introducing further challenges to 802.1Qbv scheduling. Existing solutions, using either complex optimization approaches or non-trivial scheduling heuristics, address this by scheduling frame transmissions only once they are guaranteed to have been fully received, even in the presence of clock drifts. However, this approach introduces additional delays that can impact deadline requirements. This study analytically derives tight end-to-end latency bounds, allowing us to determine if stream deadlines for a given network will be violated without the need to solve for any scheduling algorithms. It also proposes an approach that results in tighter bounds based on information collected from the synchronization process. The analytical results are compared with simulation results, confirming their validity.
- 27 November 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Networking Letters
- Vol. 7 (1) , 41-45
Open AccessLatency Bounds for TSN Scheduling in the Presence of Clock Synchronization
The IEEE 802.1Qbv (80.21Qbv) standard is designed for traffic requiring deterministic and bounded latencies through strict periodic time synchronization, as specified by IEEE 802.1AS standard. However, internal clock drift in devices causes timing misalignment, introducing further challenges to 802.1Qbv scheduling. Existing solutions, using either complex optimization approaches or non-trivial scheduling heuristics, address this by scheduling frame transmissions only once they are guaranteed to have been fully received, even in the presence of clock drifts. However, this approach introduces additional delays that can impact deadline requirements. This study analytically derives tight end-to-end latency bounds, allowing us to determine if stream deadlines for a given network will be violated without the need to solve for any scheduling algorithms. It also proposes an approach that results in tighter bounds based on information collected from the synchronization process. The analytical results are compared with simulation results, confirming their validity.
- Open Access
Multi-Stream TSN Gate Control Scheduling in the Presence of Clock Synchronization
With the advancement of technologies like Industry 4.0, communication networks must meet stringent requirements of applications demanding deterministic and bounded latencies. The problem is further compounded by the need to periodically synchronize network devices to a common time reference to address clock drifts. Existing solutions often simplify the problem by assuming either perfect synchronization or a worst-case error. Additionally, these approaches delay the scheduling process in network devices until the scheduled frame is guaranteed to have arrived in the device queue, inducing additional delays to the stream. A novel approach that completely avoids queuing delays is proposed, enabling it to meet even the strictest deadline requirement. Furthermore, both approaches can be enhanced by incorporating network-derived time-synchronization information. This is not only convenient for meeting deadline requirements but also improves bandwidth efficiency.
Open AccessMulti-Stream TSN Gate Control Scheduling in the Presence of Clock Synchronization
With the advancement of technologies like Industry 4.0, communication networks must meet stringent requirements of applications demanding deterministic and bounded latencies. The problem is further compounded by the need to periodically synchronize network devices to a common time reference to address clock drifts. Existing solutions often simplify the problem by assuming either perfect synchronization or a worst-case error. Additionally, these approaches delay the scheduling process in network devices until the scheduled frame is guaranteed to have arrived in the device queue, inducing additional delays to the stream. A novel approach that completely avoids queuing delays is proposed, enabling it to meet even the strictest deadline requirement. Furthermore, both approaches can be enhanced by incorporating network-derived time-synchronization information. This is not only convenient for meeting deadline requirements but also improves bandwidth efficiency.
- 28 June 2024
- Published by Springer Nature in Arabian Journal for Science and Engineering
- Vol. 50 (10) , 7119-7137
A Hybrid Approach Combining SAW and TOPSIS for Network Interface Selection in Heterogeneous Wireless Networks
- 28 June 2024
- Published by Springer Nature in Arabian Journal for Science and Engineering
- Vol. 50 (10) , 7119-7137
A Hybrid Approach Combining SAW and TOPSIS for Network Interface Selection in Heterogeneous Wireless Networks
- 28 January 2024
- Published by Elsevier in Computer Communications
- Vol. 217, 166-182
Mobility aware and energy-efficient federated deep reinforcement learning assisted resource allocation for 5G-RAN slicing
- 28 January 2024
- Published by Elsevier in Computer Communications
- Vol. 217, 166-182
Mobility aware and energy-efficient federated deep reinforcement learning assisted resource allocation for 5G-RAN slicing