Mustafa Özden
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Mustafa Özden has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Davut Ertekin
Power Electronics Center of Research Laboratories, Electrical and Electronics Engineering Department, Bursa Technical University, 16310 Bursa, Türkiye
4 shared publications - Nurettin Gökhan Adar
Bursa Teknik Üniversitesi
2 shared publications - Pierluigi Siano
Department of Power Electronics, Electrical Drives and Robotics of Silesian University of Technology, 44-100 Gliwice, Poland
2 shared publications - Ahmet Korkmaz1 shared publication
Canberk Şahin
Ermaksan Bursa, Trkiye
1 shared publication- Hakan Ulker
Bursa Teknik Üniversitesi
1 shared publication
Latest Publications
- Open Access
- 12 August 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (32) , 326303
Adaptive A*/NSGA-II framework for multi-goal navigation of mobile robots
The autonomous execution of routine sequential tasks by mobile robots in obstacle-dense environments poses substantial challenges for long-duration operations, as it simultaneously demands energy efficiency, kinematically feasible trajectory generation, and collision-free maneuvering. Although genetic algorithms (GAs) have been successfully applied to such problems, they often struggle to converge toward a feasible solution space when initialized with random coordinates in complex multi-goal maps. To address these limitations, we propose a hierarchical path planning framework for multi-goal navigation. Population initialization is accelerated by integrating the deterministic seed path planning of A* with the multi-objective optimization capabilities of the Non-dominated Sorting GA II (NSGA-II). Subsequently, the following five conflicting criteria are simultaneously optimized over a B-Spline curve representation to ensure that the planned paths comply with the robot’s physical constraints: path length, cumulative wheel effort (energy consumption), trajectory curvature (smoothness), obstacle clearance (centering), and collision risk. We further introduce a softmin-based adaptive blending mechanism that automatically selects the most suitable path from the champion solutions on the Pareto front according to regional environmental conditions, thereby eliminating the need for manual user intervention. Extensive simulation results across five difficulty levels involving multiple segments demonstrate that the proposed adaptive NSGA-II method achieves path length improvements of 22%–42%, 8%–27%, and 16%–34% compared with Standard GA, ant-based whale optimization algorithm-artificial potential field, and HWPSO, respectively. The adaptive blending mechanism also attains the shortest mean path length in all five tested environments.
- 12 August 2026
- Published by IOP Publishing in Measurement Science and Technology
- Vol. 37 (32) , 326303
Open AccessAdaptive A*/NSGA-II framework for multi-goal navigation of mobile robots
The autonomous execution of routine sequential tasks by mobile robots in obstacle-dense environments poses substantial challenges for long-duration operations, as it simultaneously demands energy efficiency, kinematically feasible trajectory generation, and collision-free maneuvering. Although genetic algorithms (GAs) have been successfully applied to such problems, they often struggle to converge toward a feasible solution space when initialized with random coordinates in complex multi-goal maps. To address these limitations, we propose a hierarchical path planning framework for multi-goal navigation. Population initialization is accelerated by integrating the deterministic seed path planning of A* with the multi-objective optimization capabilities of the Non-dominated Sorting GA II (NSGA-II). Subsequently, the following five conflicting criteria are simultaneously optimized over a B-Spline curve representation to ensure that the planned paths comply with the robot’s physical constraints: path length, cumulative wheel effort (energy consumption), trajectory curvature (smoothness), obstacle clearance (centering), and collision risk. We further introduce a softmin-based adaptive blending mechanism that automatically selects the most suitable path from the champion solutions on the Pareto front according to regional environmental conditions, thereby eliminating the need for manual user intervention. Extensive simulation results across five difficulty levels involving multiple segments demonstrate that the proposed adaptive NSGA-II method achieves path length improvements of 22%–42%, 8%–27%, and 16%–34% compared with Standard GA, ant-based whale optimization algorithm-artificial potential field, and HWPSO, respectively. The adaptive blending mechanism also attains the shortest mean path length in all five tested environments.
- 29 June 2026
- Published by Springer Nature in International Journal of Data Science and Analytics
- Vol. 22 (1)
Deep learning for CGI and visual forgery detection: a comprehensive survey
- 29 June 2026
- Published by Springer Nature in International Journal of Data Science and Analytics
- Vol. 22 (1)
Deep learning for CGI and visual forgery detection: a comprehensive survey
- 7 May 2026
- Published by Omer Halisdemir Universitesi in Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi
- Vol. 17
Evaluation of real-time controller performance through experimental analysis under disturbance in a one-DOF copter
One-degree-of-freedom copters, capable of a single rotational motion, are widely used as preliminary platforms for testing control strategies of lightweight rotary-wing aircraft due to their simplicity and low cost. In this study, the dynamic model of a one-degree-of-freedom copter was derived and linearized. The controller gains for the methods of Linear Quadratic Regulator (LQR) and state feedback controller by pole placement were calculated and tested through the simulations with a unit step input. Performances were evaluated using the Integral of Absolute Error (IAE) index, and LQR achieved superior results compared to pole placement. Furthermore, a feedback linearization (FL) controller based on the nonlinear model was designed and compared with LQR under sinusoidal input in the presence of nearly square-wave-like disturbances. The results with respect to the IAE index demonstrated that the FL controller offered better tracking and robustness, highlighting the advantage of incorporating the nonlinear effects in the controller design. The proposed controllers were implemented and validated on Acrome One-DOF copter, a real-time experimental setup platform using a myRIO-based control system.
- 7 May 2026
- Published by Omer Halisdemir Universitesi in Ömer Halisdemir Üniversitesi Mühendislik Bilimleri Dergisi
- Vol. 17
Evaluation of real-time controller performance through experimental analysis under disturbance in a one-DOF copter
One-degree-of-freedom copters, capable of a single rotational motion, are widely used as preliminary platforms for testing control strategies of lightweight rotary-wing aircraft due to their simplicity and low cost. In this study, the dynamic model of a one-degree-of-freedom copter was derived and linearized. The controller gains for the methods of Linear Quadratic Regulator (LQR) and state feedback controller by pole placement were calculated and tested through the simulations with a unit step input. Performances were evaluated using the Integral of Absolute Error (IAE) index, and LQR achieved superior results compared to pole placement. Furthermore, a feedback linearization (FL) controller based on the nonlinear model was designed and compared with LQR under sinusoidal input in the presence of nearly square-wave-like disturbances. The results with respect to the IAE index demonstrated that the FL controller offered better tracking and robustness, highlighting the advantage of incorporating the nonlinear effects in the controller design. The proposed controllers were implemented and validated on Acrome One-DOF copter, a real-time experimental setup platform using a myRIO-based control system.
- Top Cited Papers
- 13 January 2026
- Published by Wiley in Fuel Cells
- Vol. 26 (1)
A Levenberg–Marquardt Learning‐Based Artificial Neural Network Controller for Battery Charging in Hydrogen and Solar‐Powered Electric Vehicle Stations
Green energy and renewable energy sources (RESs) are between the most important topics in power, energy, and transportation and are crucial for sustainability for next generations. When a hydrogen fuel cell or solar array is used for the electric vehicle (EV), the next step is using an efficient, high power, and simple structure based power electronics converter to storage the energy of these RESs into the battery pack. The integration of artificial intelligence into the control and optimization of DC–DC power converters presents promising opportunities in improving energy management and efficiency in EV sector. This study presents a low‐input current and low voltage stress topology for application in fuel cell to battery charging systems in EVs. A current filter by forming a switched inductor cell at the input side of the converter guarantees a small ripple for input sources that enhances the longevity and long‐life of the FC stacks or solar panels. The application of the switched capacitor circuits at the input and end sides of the converter decreases the voltage stress across the semiconductor devices and enhances the mean time to failure rate of the converter that is between the most important features of a converter. The presented topology enhances the input voltage to 7 and 19 times for duty ratios equal to 0.5 and 0.8, respectively, while the switch experiences three and nine times the input voltage for the same duty ratios, which is considerable. The configuration of the diodes and capacitors in the switched capacitor, by dividing the total voltage stress, results in impressively low voltage ripples. This converter includes one power switch, which minimizes the complexity of the controller and enhances the feasibility. The converter design incorporates a three‐layer, three‐input artificial neural network structure. The regression values were 0.982 for training, 0.983 for testing, and 0.9827 overall, indicating minimal prediction error and confirming the effective training of the neural network model. The laboratory test results for power levels around 200 W have been presented and confirm the correctness of the proposed algorithm and the application of the proposed converter. To obtain 0.5 A for the load under a 350 VDC output voltage, the input source presents an average current equal to 8 A, and the switch experiences around 160 V voltage stress across the drain–source pins. Results show that Inductor L 2 has lower current stress than Inductor L 1 .
- 13 January 2026
- Published by Wiley in Fuel Cells
- Vol. 26 (1)
Top Cited PapersA Levenberg–Marquardt Learning‐Based Artificial Neural Network Controller for Battery Charging in Hydrogen and Solar‐Powered Electric Vehicle Stations
Green energy and renewable energy sources (RESs) are between the most important topics in power, energy, and transportation and are crucial for sustainability for next generations. When a hydrogen fuel cell or solar array is used for the electric vehicle (EV), the next step is using an efficient, high power, and simple structure based power electronics converter to storage the energy of these RESs into the battery pack. The integration of artificial intelligence into the control and optimization of DC–DC power converters presents promising opportunities in improving energy management and efficiency in EV sector. This study presents a low‐input current and low voltage stress topology for application in fuel cell to battery charging systems in EVs. A current filter by forming a switched inductor cell at the input side of the converter guarantees a small ripple for input sources that enhances the longevity and long‐life of the FC stacks or solar panels. The application of the switched capacitor circuits at the input and end sides of the converter decreases the voltage stress across the semiconductor devices and enhances the mean time to failure rate of the converter that is between the most important features of a converter. The presented topology enhances the input voltage to 7 and 19 times for duty ratios equal to 0.5 and 0.8, respectively, while the switch experiences three and nine times the input voltage for the same duty ratios, which is considerable. The configuration of the diodes and capacitors in the switched capacitor, by dividing the total voltage stress, results in impressively low voltage ripples. This converter includes one power switch, which minimizes the complexity of the controller and enhances the feasibility. The converter design incorporates a three‐layer, three‐input artificial neural network structure. The regression values were 0.982 for training, 0.983 for testing, and 0.9827 overall, indicating minimal prediction error and confirming the effective training of the neural network model. The laboratory test results for power levels around 200 W have been presented and confirm the correctness of the proposed algorithm and the application of the proposed converter. To obtain 0.5 A for the load under a 350 VDC output voltage, the input source presents an average current equal to 8 A, and the switch experiences around 160 V voltage stress across the drain–source pins. Results show that Inductor L 2 has lower current stress than Inductor L 1 .
- 30 September 2025
- Published by Hitit University in Hittite Journal of Science and Engineering
- Vol. 12 (3) , 111-120
Design and Vision-Based Control of a Low-Cost SCARA Robot
SCARA robots are widely used in industrial automation due to their high precision and speed, particularly in pick-and-place operations. In addition to conventional programming approaches, alternative vision-based control methods have gained interest to enhance flexibility and efficiency in robotic applications. This study presents the design and implementation of a Position-Based Visual Servoing (PBVS) for the SCARA robot system capable of detecting and manipulating objects in real-time. The proposed system consists of a fixed overhead camera, a SCARA robot, and Python-based control software. The software integrates image processing algorithms, kinematic calculations, and motor control, enabling the robot to autonomously identify objects, compute their positions, and execute pick and place tasks. To enhance object detection accuracy, Kuwahara filtering, Canny edge detection, morphological transformations, and connected component analysis were applied. Experimental results demonstrated that the combination of Kuwahara filtering and Canny edge detection achieved the lowest MSE error (8.45%), ensuring precise object localization. Furthermore, inverse kinematics was employed to generate accurate joint movements, allowing smooth and reliable grasping operations. The system was tested through 100 pick-and-place trials, achieving a 100% grasping success rate when Kuwahara filtering was applied. The experimental findings confirm that vision-based control significantly improves SCARA robot performance, making it suitable for automated assembly, material handling, and quality control applications.
- 30 September 2025
- Published by Hitit University in Hittite Journal of Science and Engineering
- Vol. 12 (3) , 111-120
Design and Vision-Based Control of a Low-Cost SCARA Robot
SCARA robots are widely used in industrial automation due to their high precision and speed, particularly in pick-and-place operations. In addition to conventional programming approaches, alternative vision-based control methods have gained interest to enhance flexibility and efficiency in robotic applications. This study presents the design and implementation of a Position-Based Visual Servoing (PBVS) for the SCARA robot system capable of detecting and manipulating objects in real-time. The proposed system consists of a fixed overhead camera, a SCARA robot, and Python-based control software. The software integrates image processing algorithms, kinematic calculations, and motor control, enabling the robot to autonomously identify objects, compute their positions, and execute pick and place tasks. To enhance object detection accuracy, Kuwahara filtering, Canny edge detection, morphological transformations, and connected component analysis were applied. Experimental results demonstrated that the combination of Kuwahara filtering and Canny edge detection achieved the lowest MSE error (8.45%), ensuring precise object localization. Furthermore, inverse kinematics was employed to generate accurate joint movements, allowing smooth and reliable grasping operations. The system was tested through 100 pick-and-place trials, achieving a 100% grasping success rate when Kuwahara filtering was applied. The experimental findings confirm that vision-based control significantly improves SCARA robot performance, making it suitable for automated assembly, material handling, and quality control applications.
- Open Access
- 22 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
A Comparative Study for Localization of Forgery Regions in Images
As computer technologies and image processing software have advanced, it has become progressively easier to produce simple fake or forged images by altering digital images without leaving any discernible trace. There is a significant need to detect manipulated regions in images in crucial fields such as politics, law, and forensic medicine. In this study, we propose a method that combines the traditional techniques, such as Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT), with the advantages of deep learning methods to detect manipulated regions in forged images. The proposed method involves designing an architecture where DWT and DCT are used in parallel with DenseNet based Convolutional Neural Network (CNN). To evaluate the effectiveness of this method, we implemented three alternative approaches: one that uses only DCT and CNN, another that uses only DWT and CNN, and a third that employs only CNN without either transformation. In total, four different methods were tested on eight datasets, and their performance was compared using metrics such as accuracy, precision, recall, dice similarity coefficient, and F1 score. The results from these comparisons clearly indicate the effectiveness and high classification accuracy of the proposed method. By leveraging the combined strengths of traditional image processing techniques and advanced deep learning algorithms, the proposed method demonstrates superior capability in detecting manipulated regions in forged images, thus offering a robust solution for applications in forensic field.
- 22 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. PP, 1
Open AccessA Comparative Study for Localization of Forgery Regions in Images
As computer technologies and image processing software have advanced, it has become progressively easier to produce simple fake or forged images by altering digital images without leaving any discernible trace. There is a significant need to detect manipulated regions in images in crucial fields such as politics, law, and forensic medicine. In this study, we propose a method that combines the traditional techniques, such as Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT), with the advantages of deep learning methods to detect manipulated regions in forged images. The proposed method involves designing an architecture where DWT and DCT are used in parallel with DenseNet based Convolutional Neural Network (CNN). To evaluate the effectiveness of this method, we implemented three alternative approaches: one that uses only DCT and CNN, another that uses only DWT and CNN, and a third that employs only CNN without either transformation. In total, four different methods were tested on eight datasets, and their performance was compared using metrics such as accuracy, precision, recall, dice similarity coefficient, and F1 score. The results from these comparisons clearly indicate the effectiveness and high classification accuracy of the proposed method. By leveraging the combined strengths of traditional image processing techniques and advanced deep learning algorithms, the proposed method demonstrates superior capability in detecting manipulated regions in forged images, thus offering a robust solution for applications in forensic field.
- Open Access
- 30 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 3613-3631
Levenberg-Marquardt Algorithm-Based Neural Network Smart Control Strategy for a Low-Input Current Ripple and High-Voltage Gain Power Converter in Fuel-Cells Energy Systems
A crucial aspect of DC-DC converters employed in renewable energy sources such as fuel cells is their ability to exhibit substantial increases in DC voltage and maintain an efficient structure while minimizing input current ripple. These factors play a pivotal role in enhancing the longevity of these energy sources and ensuring their compatibility with high-voltage AC and DC grids. This study introduces a high-gain DC-DC step-up converter that incorporates a continuous input current cell and a switched capacitor voltage-boosting output cell to address these requirements. The control process of this proposed converter is executed using an artificial neural network based on the Levenberg-Marquardt learning algorithm. The primary difference in this research lies in obtaining the artificial neural network-based controller directly from the circuit’s characteristic equations, rather than generating it through another controller. A feedback control strategy has been formulated, where the artificial neural network consistently produces duty increment values based on the reference voltage. Additionally, the network’s input includes not only the reference signal but also the circuit input voltage and output current value. As a result, the stability of the circuit’s output voltage is maintained against variations in input voltage and load changes. A laboratory-designed workbench underwent testing, and the experimental results substantiated the theoretical inquiries and simulation outcomes.
- 30 December 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 3613-3631
Open AccessLevenberg-Marquardt Algorithm-Based Neural Network Smart Control Strategy for a Low-Input Current Ripple and High-Voltage Gain Power Converter in Fuel-Cells Energy Systems
A crucial aspect of DC-DC converters employed in renewable energy sources such as fuel cells is their ability to exhibit substantial increases in DC voltage and maintain an efficient structure while minimizing input current ripple. These factors play a pivotal role in enhancing the longevity of these energy sources and ensuring their compatibility with high-voltage AC and DC grids. This study introduces a high-gain DC-DC step-up converter that incorporates a continuous input current cell and a switched capacitor voltage-boosting output cell to address these requirements. The control process of this proposed converter is executed using an artificial neural network based on the Levenberg-Marquardt learning algorithm. The primary difference in this research lies in obtaining the artificial neural network-based controller directly from the circuit’s characteristic equations, rather than generating it through another controller. A feedback control strategy has been formulated, where the artificial neural network consistently produces duty increment values based on the reference voltage. Additionally, the network’s input includes not only the reference signal but also the circuit input voltage and output current value. As a result, the stability of the circuit’s output voltage is maintained against variations in input voltage and load changes. A laboratory-designed workbench underwent testing, and the experimental results substantiated the theoretical inquiries and simulation outcomes.
- Open Access
- 2 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 123489-123513
The Design and Practical Realization of an Adaptable Grid Integrating Hydrogen Fuel Cell Setup With a Fuzzy-Logical Controller-Based SVPWM Boosted Inverter
The primary and fundamental requirement for a fuel cell (FC) stack is its reliable operation under various operating conditions. When FC stacks are used as the input voltage source with high ripple currents, the overall lifespan of the FC system decreases. Hence, power converter configurations need to minimize the current ripples originating from these sources. Additionally, the generated voltage from the FC stack is often lower than the required voltage level for grid connection. This paper presents a fuzzy logic controller (FLC)-equipped high-gain single-switched DC-DC boost converter. The proposed power converter topology utilizes an improved switched inductor and switched capacitor configuration to minimize input current ripples and enhance the voltage gain. The switched inductor cell is designed in such a way that its inductors charge and discharge simultaneously, effectively minimizing the input current ripple. Additionally, the proposed DC-DC boost converter utilizes a switched capacitor cell to double the generated voltage. The FLC offers real-time visualization and digital signal processing capabilities, and it is compatible with MATLAB software. For grid connection purposes, a space vector pulse width modulation (SVPWM)-based switching system is recommended, utilizing a full bridge inverter. The SVPWM technique is implemented by representing the desired output voltage with an equivalent vector VREF rotating counterclockwise, integrated with a digital signal processing (DSP)-based controller. The DSP microcontroller employed in this study operates at an 80 Mb/sec sampling speed and offers several advantages, including the ability to perform complex calculations, implement advanced control algorithms, and process signals in real-time. These capabilities contribute to enhanced performance, efficiency, and accuracy. Laboratory studies have been conducted to validate the accuracy and effectiveness of the theoretical investigations.
- 2 September 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 12, 123489-123513
Open AccessThe Design and Practical Realization of an Adaptable Grid Integrating Hydrogen Fuel Cell Setup With a Fuzzy-Logical Controller-Based SVPWM Boosted Inverter
The primary and fundamental requirement for a fuel cell (FC) stack is its reliable operation under various operating conditions. When FC stacks are used as the input voltage source with high ripple currents, the overall lifespan of the FC system decreases. Hence, power converter configurations need to minimize the current ripples originating from these sources. Additionally, the generated voltage from the FC stack is often lower than the required voltage level for grid connection. This paper presents a fuzzy logic controller (FLC)-equipped high-gain single-switched DC-DC boost converter. The proposed power converter topology utilizes an improved switched inductor and switched capacitor configuration to minimize input current ripples and enhance the voltage gain. The switched inductor cell is designed in such a way that its inductors charge and discharge simultaneously, effectively minimizing the input current ripple. Additionally, the proposed DC-DC boost converter utilizes a switched capacitor cell to double the generated voltage. The FLC offers real-time visualization and digital signal processing capabilities, and it is compatible with MATLAB software. For grid connection purposes, a space vector pulse width modulation (SVPWM)-based switching system is recommended, utilizing a full bridge inverter. The SVPWM technique is implemented by representing the desired output voltage with an equivalent vector VREF rotating counterclockwise, integrated with a digital signal processing (DSP)-based controller. The DSP microcontroller employed in this study operates at an 80 Mb/sec sampling speed and offers several advantages, including the ability to perform complex calculations, implement advanced control algorithms, and process signals in real-time. These capabilities contribute to enhanced performance, efficiency, and accuracy. Laboratory studies have been conducted to validate the accuracy and effectiveness of the theoretical investigations.