This paper presents simulation and experimental validation of a Nonlinear Compensation-based Neuro Fuzzy (NCNF) controller designed to balance the rotary inverted pendulum (RIP). Traditional linear controllers, such as Proportional-Integral-Derivative (PID) and state-feedback with pole placement, usually achieve satisfactory results in simulations on linearized models. However, their performance decreases in hardware implementation because of disturbances and unmodeled nonlinear effects such as Coulomb friction and mechanical backlash. To overcome these challenges, a feedforward compensation function was developed to cancel these undesired effects, which is combined with an Adaptive Neuro-Fuzzy Inference System (ANFIS) controller that updates PID gains to improve the rotary arm tracking for a square-wave reference and stabilize the pendulum at the upright position. The proposed NCNF controller is validated through hardware-in-the-loop (HIL) experiments and compared with a baseline state-feedback controller. Results show that the arm angle (θ) overshoot decreased from 40.6% to 0.8% (lower step) and from 17.2% to 2.5% (upper), total steady-state θ-error from 5.75° to 0.296°, and the fitness index dropped from 41.12 to 25.23. The nonlinear compensation reduced the gap between simulation and real-time performance, while the ANFIS further improved the defined control metrics. Overall, the NCNF controller achieves more stable and precise tracking than the state-feedback control.

This paper presents simulation and experimental validation of a Nonlinear Compensation-based Neuro Fuzzy (NCNF) controller designed to balance the rotary inverted pendulum (RIP). Traditional linear controllers, such as Proportional-Integral-Derivative (PID) and state-feedback with pole placement, usually achieve satisfactory results in simulations on linearized models. However, their performance decreases in hardware implementation because of disturbances and unmodeled nonlinear effects such as Coulomb friction and mechanical backlash. To overcome these challenges, a feedforward compensation function was developed to cancel these undesired effects, which is combined with an Adaptive Neuro-Fuzzy Inference System (ANFIS) controller that updates PID gains to improve the rotary arm tracking for a square-wave reference and stabilize the pendulum at the upright position. The proposed NCNF controller is validated through hardware-in-the-loop (HIL) experiments and compared with a baseline state-feedback controller. Results show that the arm angle (θ) overshoot decreased from 40.6% to 0.8% (lower step) and from 17.2% to 2.5% (upper), total steady-state θ-error from 5.75° to 0.296°, and the fitness index dropped from 41.12 to 25.23. The nonlinear compensation reduced the gap between simulation and real-time performance, while the ANFIS further improved the defined control metrics. Overall, the NCNF controller achieves more stable and precise tracking than the state-feedback control.

This article focuses on improving fault tolerance and optimizing energy consumption in the context of a mining prospection mission conducted by a swarm of autonomous robots. Two major contributions are proposed. The first aims to reduce communication between robots in order to increase the system’s robustness in the presence of failures. The second focuses on minimizing the trajectory of a deminer robot to reduce overall energy consumption. To address these goals, two reinforcement learning based algorithms are proposed: Deep Q-Network (DQN) and Federated Reinforcement Learning (FRL), both derived from the Q-learning algorithm. Simulation results examining the impact of the exploration rate α on the number of detected mines show that, with 10 autonomous robots of the same architecture and 30 randomly placed mines over 30 experiments, the FRL algorithm provides better fault tolerance and ensures that the main prospection mission is accomplished even in the presence of some robotic failures or errors. Furthermore, a second series of 60 experiments involving the integration of the deminer robot, focused on optimizing energy consumption, demonstrates that the DQN algorithm is more effective in reducing energy usage, due to improved a better optimization of unnecessary deminer movements, while successfully resolving deadlock situations that the latter may encounter. These findings open the door to the development of a hybrid algorithm combining the strengths of DQN and FRL to ensure both system robustness and minimal energy consumption.

This article focuses on improving fault tolerance and optimizing energy consumption in the context of a mining prospection mission conducted by a swarm of autonomous robots. Two major contributions are proposed. The first aims to reduce communication between robots in order to increase the system’s robustness in the presence of failures. The second focuses on minimizing the trajectory of a deminer robot to reduce overall energy consumption. To address these goals, two reinforcement learning based algorithms are proposed: Deep Q-Network (DQN) and Federated Reinforcement Learning (FRL), both derived from the Q-learning algorithm. Simulation results examining the impact of the exploration rate α on the number of detected mines show that, with 10 autonomous robots of the same architecture and 30 randomly placed mines over 30 experiments, the FRL algorithm provides better fault tolerance and ensures that the main prospection mission is accomplished even in the presence of some robotic failures or errors. Furthermore, a second series of 60 experiments involving the integration of the deminer robot, focused on optimizing energy consumption, demonstrates that the DQN algorithm is more effective in reducing energy usage, due to improved a better optimization of unnecessary deminer movements, while successfully resolving deadlock situations that the latter may encounter. These findings open the door to the development of a hybrid algorithm combining the strengths of DQN and FRL to ensure both system robustness and minimal energy consumption.

This article focuses on improving fault tolerance and optimizing energy consumption in the context of a mining prospection mission conducted by a swarm of autonomous robots. Two major contributions are proposed. The first aims to reduce communication between robots in order to increase the system’s robustness in the presence of failures. The second focuses on minimizing the trajectory of a deminer robot to reduce overall energy consumption. To address these goals, two reinforcement learning based algorithms are proposed: Deep Q-Network (DQN) and Federated Reinforcement Learning (FRL), both derived from the Q-learning algorithm. Simulation results examining the impact of the exploration rate α on the number of detected mines show that, with 10 autonomous robots of the same architecture and 30 randomly placed mines over 30 experiments, the FRL algorithm provides better fault tolerance and ensures that the main prospection mission is accomplished even in the presence of some robotic failures or errors. Furthermore, a second series of 60 experiments involving the integration of the deminer robot, focused on optimizing energy consumption, demonstrates that the DQN algorithm is more effective in reducing energy usage, due to improved a better optimization of unnecessary deminer movements, while successfully resolving deadlock situations that the latter may encounter. These findings open the door to the development of a hybrid algorithm combining the strengths of DQN and FRL to ensure both system robustness and minimal energy consumption.

This article focuses on improving fault tolerance and optimizing energy consumption in the context of a mining prospection mission conducted by a swarm of autonomous robots. Two major contributions are proposed. The first aims to reduce communication between robots in order to increase the system’s robustness in the presence of failures. The second focuses on minimizing the trajectory of a deminer robot to reduce overall energy consumption. To address these goals, two reinforcement learning based algorithms are proposed: Deep Q-Network (DQN) and Federated Reinforcement Learning (FRL), both derived from the Q-learning algorithm. Simulation results examining the impact of the exploration rate α on the number of detected mines show that, with 10 autonomous robots of the same architecture and 30 randomly placed mines over 30 experiments, the FRL algorithm provides better fault tolerance and ensures that the main prospection mission is accomplished even in the presence of some robotic failures or errors. Furthermore, a second series of 60 experiments involving the integration of the deminer robot, focused on optimizing energy consumption, demonstrates that the DQN algorithm is more effective in reducing energy usage, due to improved a better optimization of unnecessary deminer movements, while successfully resolving deadlock situations that the latter may encounter. These findings open the door to the development of a hybrid algorithm combining the strengths of DQN and FRL to ensure both system robustness and minimal energy consumption.

Designing scalable and interpretable control strategies for decentralized multi-agent systems remains a challenge in reinforcement learning (RL). This challenge is particularly evident in pursuit–evasion tasks, which require coordination under partial observability, without explicit communication or centralized guidance. Although deep RL methods achieve strong performance, they typically operate as black boxes, limiting trust and deployment in safety-critical domains. We propose a Multi-Head DDPG architecture that decomposes control into three interpretable force components - pursuit, cohesion, and separation - weighted adaptively to generate context-aware actions. This design enables emergent role differentiation and interpretable self-organization in the model. In grid-based pursuit–evasion benchmarks, our method outperforms DQN, PPO, and standard DDPG in terms of success rate, convergence speed, and generalization, while also yielding transparent collective behaviors. Overall, the results show that weighted force-based behavioral decomposition provides a principled pathway toward achieving both high-performance and explainable multi-agent control.

Designing scalable and interpretable control strategies for decentralized multi-agent systems remains a challenge in reinforcement learning (RL). This challenge is particularly evident in pursuit–evasion tasks, which require coordination under partial observability, without explicit communication or centralized guidance. Although deep RL methods achieve strong performance, they typically operate as black boxes, limiting trust and deployment in safety-critical domains. We propose a Multi-Head DDPG architecture that decomposes control into three interpretable force components - pursuit, cohesion, and separation - weighted adaptively to generate context-aware actions. This design enables emergent role differentiation and interpretable self-organization in the model. In grid-based pursuit–evasion benchmarks, our method outperforms DQN, PPO, and standard DDPG in terms of success rate, convergence speed, and generalization, while also yielding transparent collective behaviors. Overall, the results show that weighted force-based behavioral decomposition provides a principled pathway toward achieving both high-performance and explainable multi-agent control.

Designing scalable and interpretable control strategies for decentralized multi-agent systems remains a challenge in reinforcement learning (RL). This challenge is particularly evident in pursuit–evasion tasks, which require coordination under partial observability, without explicit communication or centralized guidance. Although deep RL methods achieve strong performance, they typically operate as black boxes, limiting trust and deployment in safety-critical domains. We propose a Multi-Head DDPG architecture that decomposes control into three interpretable force components - pursuit, cohesion, and separation - weighted adaptively to generate context-aware actions. This design enables emergent role differentiation and interpretable self-organization in the model. In grid-based pursuit–evasion benchmarks, our method outperforms DQN, PPO, and standard DDPG in terms of success rate, convergence speed, and generalization, while also yielding transparent collective behaviors. Overall, the results show that weighted force-based behavioral decomposition provides a principled pathway toward achieving both high-performance and explainable multi-agent control.

Designing scalable and interpretable control strategies for decentralized multi-agent systems remains a challenge in reinforcement learning (RL). This challenge is particularly evident in pursuit–evasion tasks, which require coordination under partial observability, without explicit communication or centralized guidance. Although deep RL methods achieve strong performance, they typically operate as black boxes, limiting trust and deployment in safety-critical domains. We propose a Multi-Head DDPG architecture that decomposes control into three interpretable force components - pursuit, cohesion, and separation - weighted adaptively to generate context-aware actions. This design enables emergent role differentiation and interpretable self-organization in the model. In grid-based pursuit–evasion benchmarks, our method outperforms DQN, PPO, and standard DDPG in terms of success rate, convergence speed, and generalization, while also yielding transparent collective behaviors. Overall, the results show that weighted force-based behavioral decomposition provides a principled pathway toward achieving both high-performance and explainable multi-agent control.

Designing scalable and interpretable control strategies for decentralized multi-agent systems remains a challenge in reinforcement learning (RL). This challenge is particularly evident in pursuit–evasion tasks, which require coordination under partial observability, without explicit communication or centralized guidance. Although deep RL methods achieve strong performance, they typically operate as black boxes, limiting trust and deployment in safety-critical domains. We propose a Multi-Head DDPG architecture that decomposes control into three interpretable force components - pursuit, cohesion, and separation - weighted adaptively to generate context-aware actions. This design enables emergent role differentiation and interpretable self-organization in the model. In grid-based pursuit–evasion benchmarks, our method outperforms DQN, PPO, and standard DDPG in terms of success rate, convergence speed, and generalization, while also yielding transparent collective behaviors. Overall, the results show that weighted force-based behavioral decomposition provides a principled pathway toward achieving both high-performance and explainable multi-agent control.

Designing scalable and interpretable control strategies for decentralized multi-agent systems remains a challenge in reinforcement learning (RL). This challenge is particularly evident in pursuit–evasion tasks, which require coordination under partial observability, without explicit communication or centralized guidance. Although deep RL methods achieve strong performance, they typically operate as black boxes, limiting trust and deployment in safety-critical domains. We propose a Multi-Head DDPG architecture that decomposes control into three interpretable force components - pursuit, cohesion, and separation - weighted adaptively to generate context-aware actions. This design enables emergent role differentiation and interpretable self-organization in the model. In grid-based pursuit–evasion benchmarks, our method outperforms DQN, PPO, and standard DDPG in terms of success rate, convergence speed, and generalization, while also yielding transparent collective behaviors. Overall, the results show that weighted force-based behavioral decomposition provides a principled pathway toward achieving both high-performance and explainable multi-agent control.

In this study, a numerical investigation was conducted on the seismic behavior of low-strength reinforced concrete columns, strengthened with steel bars and wrapped with fiberglass tapes and fabrics, using finite element software. The columns were subjected to both monotonic and cyclic loading, and the analysis focused on fracture patterns, failure mechanisms, lateral hysteresis loops, ductility degradation, and stiffness degradation. The results showed that the reference column exhibited brittle shear failure and insufficient ductility. In contrast, the second column, reinforced with steel bars and partially wrapped with fiberglass tapes, demonstrated 30% higher tensile strength compared to the reference column, achieving stable hysteresis loops, improved energy dissipation, and 25% less cracking. The third column, fully wrapped with fiberglass fabric in addition to the steel bars, exhibited 50% higher tensile strength and 75% reduced probability of cracking in the plastic hinge area. These findings underscore the effectiveness of advanced reinforcement techniques in improving the seismic performance of reinforced concrete columns.

In this study, a numerical investigation was conducted on the seismic behavior of low-strength reinforced concrete columns, strengthened with steel bars and wrapped with fiberglass tapes and fabrics, using finite element software. The columns were subjected to both monotonic and cyclic loading, and the analysis focused on fracture patterns, failure mechanisms, lateral hysteresis loops, ductility degradation, and stiffness degradation. The results showed that the reference column exhibited brittle shear failure and insufficient ductility. In contrast, the second column, reinforced with steel bars and partially wrapped with fiberglass tapes, demonstrated 30% higher tensile strength compared to the reference column, achieving stable hysteresis loops, improved energy dissipation, and 25% less cracking. The third column, fully wrapped with fiberglass fabric in addition to the steel bars, exhibited 50% higher tensile strength and 75% reduced probability of cracking in the plastic hinge area. These findings underscore the effectiveness of advanced reinforcement techniques in improving the seismic performance of reinforced concrete columns.

In this study, a numerical investigation was conducted on the seismic behavior of low-strength reinforced concrete columns, strengthened with steel bars and wrapped with fiberglass tapes and fabrics, using finite element software. The columns were subjected to both monotonic and cyclic loading, and the analysis focused on fracture patterns, failure mechanisms, lateral hysteresis loops, ductility degradation, and stiffness degradation. The results showed that the reference column exhibited brittle shear failure and insufficient ductility. In contrast, the second column, reinforced with steel bars and partially wrapped with fiberglass tapes, demonstrated 30% higher tensile strength compared to the reference column, achieving stable hysteresis loops, improved energy dissipation, and 25% less cracking. The third column, fully wrapped with fiberglass fabric in addition to the steel bars, exhibited 50% higher tensile strength and 75% reduced probability of cracking in the plastic hinge area. These findings underscore the effectiveness of advanced reinforcement techniques in improving the seismic performance of reinforced concrete columns.

In this study, a numerical investigation was conducted on the seismic behavior of low-strength reinforced concrete columns, strengthened with steel bars and wrapped with fiberglass tapes and fabrics, using finite element software. The columns were subjected to both monotonic and cyclic loading, and the analysis focused on fracture patterns, failure mechanisms, lateral hysteresis loops, ductility degradation, and stiffness degradation. The results showed that the reference column exhibited brittle shear failure and insufficient ductility. In contrast, the second column, reinforced with steel bars and partially wrapped with fiberglass tapes, demonstrated 30% higher tensile strength compared to the reference column, achieving stable hysteresis loops, improved energy dissipation, and 25% less cracking. The third column, fully wrapped with fiberglass fabric in addition to the steel bars, exhibited 50% higher tensile strength and 75% reduced probability of cracking in the plastic hinge area. These findings underscore the effectiveness of advanced reinforcement techniques in improving the seismic performance of reinforced concrete columns.

The automated inspection of complex freeform components, such as helical gears, is a persistent challenge in advanced manufacturing due to their intricate geometries and strict precision requirements. Conventional inspection methods, such as those using coordinate measuring machines or optical techniques, are often time-consuming and lack adaptability to subtle deviations. Recent deep learning approaches show promise but are typically limited to point-based or scan-to-scan comparisons, which remain sensitive to noise and alignment errors. We propose a voxel-based 3D inspection framework that integrates an XGBoost-guided perturbation model with a 3D convolutional autoencoder (3D CNN-AE). CAD-derived gear models are systematically perturbed with controlled Gaussian deformations to emulate tolerances, defects, and sensor noise, then voxelized for autoencoder training. This enables robust learning of nominal gear geometry distributions. Extensive experiments conducted against PointNet++, a Variational Autoencoder, and a GAN-based reconstruction model demonstrate that our method consistently achieves superior performance across various metrics, including PSNR, SSIM, accuracy, precision, recall, and F1-score. The results highlight the potential of voxel-based learning with data-driven perturbation for scalable and high-accuracy inspection in industrial applications.

The automated inspection of complex freeform components, such as helical gears, is a persistent challenge in advanced manufacturing due to their intricate geometries and strict precision requirements. Conventional inspection methods, such as those using coordinate measuring machines or optical techniques, are often time-consuming and lack adaptability to subtle deviations. Recent deep learning approaches show promise but are typically limited to point-based or scan-to-scan comparisons, which remain sensitive to noise and alignment errors. We propose a voxel-based 3D inspection framework that integrates an XGBoost-guided perturbation model with a 3D convolutional autoencoder (3D CNN-AE). CAD-derived gear models are systematically perturbed with controlled Gaussian deformations to emulate tolerances, defects, and sensor noise, then voxelized for autoencoder training. This enables robust learning of nominal gear geometry distributions. Extensive experiments conducted against PointNet++, a Variational Autoencoder, and a GAN-based reconstruction model demonstrate that our method consistently achieves superior performance across various metrics, including PSNR, SSIM, accuracy, precision, recall, and F1-score. The results highlight the potential of voxel-based learning with data-driven perturbation for scalable and high-accuracy inspection in industrial applications.

The automated inspection of complex freeform components, such as helical gears, is a persistent challenge in advanced manufacturing due to their intricate geometries and strict precision requirements. Conventional inspection methods, such as those using coordinate measuring machines or optical techniques, are often time-consuming and lack adaptability to subtle deviations. Recent deep learning approaches show promise but are typically limited to point-based or scan-to-scan comparisons, which remain sensitive to noise and alignment errors. We propose a voxel-based 3D inspection framework that integrates an XGBoost-guided perturbation model with a 3D convolutional autoencoder (3D CNN-AE). CAD-derived gear models are systematically perturbed with controlled Gaussian deformations to emulate tolerances, defects, and sensor noise, then voxelized for autoencoder training. This enables robust learning of nominal gear geometry distributions. Extensive experiments conducted against PointNet++, a Variational Autoencoder, and a GAN-based reconstruction model demonstrate that our method consistently achieves superior performance across various metrics, including PSNR, SSIM, accuracy, precision, recall, and F1-score. The results highlight the potential of voxel-based learning with data-driven perturbation for scalable and high-accuracy inspection in industrial applications.

Oil well drilling is an expensive process that needs a particular focus. For this reason, Rate Of Penetration (ROP) has been widely approved as a measure of drilling efficiency and adequate configuration parameters. Our aim in this work consists in the elaboration of a smart system using Hoeffding trees for predicting the Rate of Penetration (ROP) in oilfield drilling. The choice of Hoeffding trees to build our model is motivated by their adaptive learning capability and drift detection. They offer continuous, fast, and efficient learning both online on data streams and offline on batch data. To validate our approach, we used real drilling data from the “Hassi-Terfa” oilfield located in Southeast Algeria. The obtained results show in comparison to the eXtreme Gradient Boosting (XGBoost) algorithm that Hoeffding trees maintain their learning capacity and produce more accurate predictions even in the presence of drifts. This is thanks to the combination of the Adaptive Windowing (ADWIN) algorithm to manage drifts and least mean squares (LMS) filters to reduce noise. This observation highlights the effectiveness of our approach to predict the ROP while oil-well drilling. The proposed smart system offers more efficient solution to predict the ROP, whether in real-time or offline. By leveraging its adaptability to changes in data distribution, our approach ensures more accurate and adaptive predictions, facilitating drilling operations optimization and boosting the overall efficiency of the process.