Indoor localization is a new technology to monitor of the health condition movements of individual patients that can be accomplished with the assistance of AI driven approach with the multimodal data fusion, i.e. geomagnetic sensors fields – accelerometers, gyroscopes, magnetometers and Wi-Fi signals. The current methods have three key limitations: (i) no robust cross-modal interaction modelling, (ii) low sensor-noise and environment-variable adaptability and (iii) no integrated health-conscious mobility tracking. The concept of finding the location of a person or item in a confined space where GPS signals cannot be used is known as indoor localization. Wi-Fi RSSI, inertial measurement units (IMU) and geomagnetic sensors have become the multimodal data fusion that has demonstrated great power in enhancing localization accuracy. This work employs two combinatorial architectures, a hierarchical clustering tree (HCT) fusion model model used for fine-grained hierarchical representation and a RF dominant ensemble used for generalization under sensor noise and variability. The hierarchical clustering tree (HCT) based approach proposed in this study can achieve a classification accuracy of 91.73%, and its macro F1 score is 0.9173 in the next-room prediction task its mean absolute error (MAE) in the travel time prediction task is 2.87 s.
Low-cost minidrone platforms are increasingly used in Unmanned Aerial Vehicle (UAV) research for safe indoor experimentation, controller development and engineering education. The Parrot Mambo Minidrone is particularly attractive because of its compact size, onboard sensing, MATLAB/Simulink compatibility and suitability for real-time validation. This review critically synthesises 137 studies published between 2018 and 2026 involving the Parrot Mambo, including studies where it appears alongside other UAV platforms. The literature spans modelling, state estimation, classical and intelligent control, robust nonlinear control, disturbance rejection, vision-based navigation, multi-UAV coordination, Human-Drone Interaction (HDI), software validation, cyber-resilience, education, safety analysis, energy-aware evaluation and hardware performance. A platform-centred taxonomy is developed to trace progression from Proportional-Integral-Derivative (PID)-based stabilisation toward adaptive, intelligent, perception-driven, software-aware and safety-conscious autonomy. PID and Proportional-Integral–Proportional-Derivative (PI-PD) controllers remain important baselines, while fuzzy-PID, Sliding Mode Control (SMC), Model Predictive Control (MPC), Model-Free Control (MFC), Reinforcement Learning (RL), Observer-Based Control (OBC) and Fault-Tolerant Control (FTC) address nonlinearities, disturbances, uncertainty and hardware constraints. However, software testing, cyber-resilience, fault diagnosis, reachability analysis, actuator health, aerodynamic constraints and energy-aware evaluation remain weakly integrated. The principal gap is the absence of a standardised benchmark jointly assessing control accuracy, robustness, perception reliability, energy use, computational load, software dependability, cyber-resilience, safety, actuator constraints and real-flight repeatability. Accordingly, this review proposes a benchmark-driven framework for integrated, safety-aware, hardware-aware and reproducible Parrot Mambo research.
Optimizing oxygen stoichiometry control in polymer electrolyte membrane fuel cell (PEMFC) systems is challenging because of load disturbances, actuator saturation, output constraints, uncertainties, and nonlinearities. This paper proposes a novel ultra-local model (ULM) that relies only on PEMFC input and output signals. Based on this model, an ultra-local model-based adaptive fixed-time fractional-order prescribed performance control (U-AFxTFOPPC) scheme is developed by integrating an adaptive fixed-time disturbance observer using a radial basis function neural network (RBFNN), an adaptive compensator, and fixed-time fractional-order prescribed performance control. The ULM reduces modeling complexity while improving flexibility and adaptability. The disturbance observer estimates the lumped disturbance and guarantees fixed-time convergence independent of initial conditions. The FxTFOPPC scheme constrains the tracking error within prescribed performance bounds while ensuring closed-loop stability. In addition, the adaptive compensator mitigates estimation errors, suppresses input chattering, and improves robustness. The stability of the proposed control strategy is verified using the Lyapunov method. Finally, MATLAB/Simulink simulations based on a nonlinear PEMFC model demonstrate effective oxygen stoichiometry regulation under load variations and actuator constraints, with enhanced tracking accuracy, robustness, and transient performance compared with existing control approaches. These results confirm its suitability for practical PEMFC air-supply control applications subject to uncertain and constrained operating conditions.
High-speed mud-pulse telemetry requires a rotary actuator that can follow prescribed phase trajectories under structured hydraulic and magnetic load torque. This paper presents a direct-drive rotary siren pulser controlled by a nested vector-control architecture augmented with a virtual-model reference channel and an angle-synchronous feedforward compensation block. The prototype uses a three-blade siren, a permanent-magnet synchronous motor (PMSM), a 4096-count encoder, and an automatically calibrated disturbance map over the full rotor angle. In the steady-state quadrature phase-shift keying (QPSK) 20 Hz carrier/20 baud regime, the virtual-model channel reduced peak absolute angular error from 120 counts (10.55 deg) to 32 counts (2.81 deg) and reduced root-mean-square (RMS) error from 45.0 to 9.0 counts. With the parking disturbance engaged, the angle-synchronous compensation reduced peak/RMS error from 340/168.1 to 44/9.73 counts at 1 Hz and from 181/66.3 to 67/23.8 counts at 20 Hz. In a one-well dataset, the proposed system demodulated in differential smooth quadrature phase-shift keying (DSQPSK) mode retained 84.3% of received bits above a 90% confidence threshold, compared with 23.3% for the comparator QPSK system. The results show that deterministic angle-dependent load torque can be compensated directly in the control architecture, improving high-speed phase tracking and receive quality without a mechanical transmission stage.
Multimodal medical image fusion (MMIF) aims to integrate complementary information from different imaging modalities into a single image, which can facilitate visual inspection and subsequent analysis. Existing deep learning (DL)-based methods mainly focus on bridging inter-modal heterogeneity, while the handling of inter-modal common information is relatively less emphasized. Inadequate treatment of such redundancy may introduce feature interference or blurring artifacts in fused results. In addition, oversimplified multi-source feature aggregation can lead to the loss of fine details. To address these issues, we propose SFCFuse, a saliency-guided fusion framework with multi-source feature interaction. Specifically, we adopt a Swin-Unet backbone for hierarchical multi-scale feature extraction and introduce a Saliency Feature Guidance (SFG) module with gated selection to suppress redundant responses and emphasize informative regions. Furthermore, a Joint Feature Interaction Fusion (JFIF) module is designed to enable attention-based interaction between heterogeneous feature sources, facilitating complementary information exchange. Experimental results on PET-MRI and SPECT-MRI datasets demonstrate that SFCFuse achieves competitive performance against representative MMIF methods in both quantitative metrics and visual comparisons.
