Complex systems, considered in control engineering, are often modeled in the form of a multi-input-multi-output (MIMO) coupled model. These systems pose definite challenges, due to their coupling behavior and the associated high computational costs. This paper describes a proposal for a technique for designing proportional-integral-derivative plus filter (PIDF) controllers for high order MIMO systems, while compensating the inherited coupling effect. The high-order MIMO system is decoupled into single-input-single- output (SISO) subsystems, which are then investigated by analyzing their time and frequency responses, pole-zero maps, and closed-loop stability. PIDF controllers with phase lead compensators are designed for each subsystem model so as to satisfy the desired closed-loop specifications. Additional controllers are introduced in order to eliminate the coupling effects in the coupled control system, while respecting the physical constraints. The proposed technique is validated through numerical simulations on a coupled control system with a high-order MIMO model, which demonstrate that the proposed approach can achieve the control requirements with acceptable performance for various reference inputs.
This research presents an innovative methodology for designing, optimizing, and implementing observer-based control systems for 3DOF hover Unmanned Aerial Vehicles (UAVs) using Lie group modeling. When addressing the challenges of nonlinear dynamics and system uncertainties, the study leverages the inherent geometric structure of Lie groups, allowing rotation and translation to be modeled naturally on SO(3) and SE(3) manifolds. This ensures geometric consistency and avoids the singularities and drift associated with conventional approaches. The proposed framework integrates a geometric observer, operating directly on the Lie group manifold with a family of controllers including PID, LQR, and Back-stepping whose parameters are systematically tuned using Multiobjective Particle Swarm Optimization (MPSO). Extensive simulations and real-time experiments on a Quanser 3-DOF Hover platform demonstrate that the proposed approach delivers substantial improvements over traditional Euclidean observer-controller schemes, achieving up to 48% reduction in estimation error, 51% faster convergence, and 16% lower energy consumption. The system also exhibits superior trajectory tracking and disturbance rejection under varied operational conditions. These results validate not only the practical feasibility but also the significant advantages of Lie group-based observer-controller design, offering a robust and energy-efficient solution for advanced UAV applications requiring high precision, resilience, and reliability.
Magnetic Resonance Imaging (MRI) is a non-invasive medical imaging technique, which is widely used for the diagnosis and treatment of various neurological disorders, including brain tumors. Accurate segmentation of brain tumors in MRI images is essential for effective diagnosis and treatment planning. Deep learning algorithms have shown significant potential in improving the speed and accuracy of medical image analysis, while reducing human error. In this paper, a deep learning-based approach called Deep Networks with Stochastic Depth (DNSD) is proposed for brain tumor segmentation in MRI images. It leverages the advantages of deep convolutional neural networks, combined with stochastic depth training to enhance feature learning and reduce training complexity. The performance of the proposed technique is evaluated using a publicly available dataset and compared with state-of-the-art deep learning models, such as U-Net and ResNet. Experimental results demonstrate that the proposed method achieves improved segmentation accuracy and computational efficiency compared to existing techniques.
In recent years, Content-Based Image Retrieval (CBIR) has advanced significantly with Deep Learning (DL), yet most of the existing models rely on global descriptors and attention-based local features, which struggle to capture explicit structural relationships within images. These models often operate as black boxes and provide limited interpretability; they exhibit reduced robustness under viewpoint and geometric variations. To address these limitations, an Interpretable Region-Attribute Graph Network (IR-AGN) is proposed in this research. Specifically, each image is decomposed into semantically meaningful regions, where each node encodes visual and semantic descriptors while edges represent geometric, appearance, and contextual relations. Consequently, a joint node-edge graph matching strategy computes similarity by aligning visual descriptors and relational structures between query and retrieval images. The results of the proposed IR-AGN model exhibit a higher mean Average Precision (mAP) of 95.19% on the Revisited Paris (RParis) dataset, when compared with the existing Multi-Layer Orientation Histogram (MLOH) model.
