https://jst.vn/index.php/ssad/issue/feed 2026-01-19T09:11:22+00:00 Open Journal Systems https://jst.vn/index.php/ssad/article/view/1193 2025-12-12T03:02:16+00:00 Hiep Nguyen-Hoang vinh.tranquang1@hust.edu.vn Quyen Nguyen-Van vinh.tranquang1@hust.edu.vn Vinh Tran-Quang vinh.tranquang1@hust.edu.vn

Air transportation is one of the fastest and safest modes of transport today. However, ensuring its safety and efficiency requires effective management of various risk factors, particularly foreign object debris (FOD) on airport runways. FOD can cause severe damage to aircraft engines and structures, disrupt operations, and even lead to fatal accidents. This study presents a method for detecting, classifying, and estimating the size of FOD on airport runways. The system integrates YOLOv11, a calibrated camera, Canny Edge Detection, and a LiDAR sensor and deploy on a Jetson Nano for efficient processing. Experimental results demonstrate the ability of the system to accurately classify FOD and measure its size in real time. The system achieves an accuracy ranging from 70% to 100% within a 0 to 3 metre distance between the FOD and the camera. This contributes to more efficient FOD detection and collection using robots or rovers, thereby enhancing runway safety and reducing risks in aviation operations.

2025-12-20T00:00:00+00:00 Copyright (c) 2026 Smart Systems and Devices https://jst.vn/index.php/ssad/article/view/1202 2025-12-18T04:19:42+00:00 Trung Nguyen Huu trung.nguyenhuu@hust.edu.vn

In the paper, we propose a novel beamforming-based Orthogonal Time Frequency Space (OTFS) transmission framework for UAV-to-Satellite Communication (U2SC) tailored for 6G-Enabled Internet of Vehicles (IoV) networks. To address the unique challenges of high Doppler shifts, long-range line-of-sight (LoS) links, and fast-moving Low Earth Orbit (LEO) satellites, we adopt OTFS modulation due to its inherent robustness against doubly dispersive channels. A Uniform Linear Array (ULA) is equipped on the UAV to enable highly directional transmission. Furthermore, we propose a Deep Q-Learning (DQL) framework for adaptive beamforming, in which the beam control problem is formulated as a Markov Decision Process (MDP). By leveraging DQL, the agent learns to dynamically steer the beam to align with the satellite’s trajectory, optimizing both link quality and energy efficiency while minimizing misalignment. Simulation results demonstrate significant gains in signal robustness and beam alignment accuracy compared to conventional methods. In addition, future work will focus on building a hardware-in-the-loop (HIL) testbed using a UAV platform with phased-array antennas to validate the proposed model under real orbital satellite trajectories and Doppler conditions.

2025-12-20T00:00:00+00:00 Copyright (c) 2026 Smart Systems and Devices https://jst.vn/index.php/ssad/article/view/1194 2025-12-12T03:37:35+00:00 Son T. Nguyen son.nguyenthanh@hust.edu.vn Tu M. Pham son.nguyenthanh@hust.edu.vn Anh Hoang anh.hoang@hust.edu.vn Trung T. Cao trung.caothanh@hust.edu.vn Quang M. Tran quang.tm212596@sis.hust.edu.vn

Gesture recognition has become an important focus in human-machine interaction (HMI). Static hand gesture recognition is particularly useful for detecting the intuitive intentions of individuals who are deaf or mute. In recent years, various data gloves have been developed to capture static hand gestures. These gloves, worn like regular gloves, serve as input devices for HMI systems. A low-cost data glove can be built using flex sensors to detect finger bending, enabling the collection of data on finger positions for different static gestures. This data can then be interpreted by a computer program. While many commercial data gloves are bundled with software, they are often prohibitively expensive. This research develops a low-cost data glove using flex sensors and an Arduino Nano. For accurate static gesture recognition, a Bayesian neural network (BNN) is employed to classify different gestures. To optimize training, the scaled-conjugate gradient method, an efficient, automated algorithm, is used to update the network’s weights and biases.

2025-12-20T00:00:00+00:00 Copyright (c) 2026 Smart Systems and Devices https://jst.vn/index.php/ssad/article/view/1195 2025-12-12T03:58:22+00:00 Quang Nguyen Hong quang.nguyenhong1@hust.edu.vn Ninh Nguyen Tuan ninh.nguyentuan@hust.edu.vn

This paper presents a novel approach to hydro turbine speed control by integrating neural networks into the tuning process of digital governors, with the objective of meeting the stringent performance requirements set by the National Power System Control Center of Vietnam Electricity (EVN). The proposed control strategy employs a closed-loop configuration using turbine rotational speed as feedback to regulate water flow and maintain power balance under varying load conditions. A key innovation of the study is the use of an adaptive neural network - proportional integral derivative (NN-PID) Controller, which continuously updates control parameters in real-time through the Brandt-Lin learning algorithm. This allows the controller to respond effectively to nonlinearities and disturbances in the system. Simulation and hardware-in-the-loop experiments validate the effectiveness of the proposed method, demonstrating enhanced performance compared to traditional proportional integral derivative (PID) controllers including faster settling times, zero steady-state error, and suppression of oscillations during sudden load changes. The results suggest that the NN-PID controller offers a promising alternative for next-generation digital governor design in large hydropower plants.

2025-12-20T00:00:00+00:00 Copyright (c) 2026 Smart Systems and Devices https://jst.vn/index.php/ssad/article/view/1197 2025-12-12T08:12:07+00:00 Huu Cong Vu phuong.vuhoang@hust.edu.vn Tuan Dau Quang phuong.vuhoang@hust.edu.vn Tung Dao Do Duy phuong.vuhoang@hust.edu.vn Phuong Vu Hoang phuong.vuhoang@hust.edu.vn

In the paper, a new three-phase five-level voltage source inverter (VSI) topology is developed by utilizing six three-level F-type legs. The proposed topology reduces the number of power electronic components while operating with a single DC source, offering advantages over conventional three-phase five-level VSI topologies. The operating principles are analyzed theoretically, and a simplified space vector modulation (SVM) method is introduced to control the proposed topology. In the proposed SVM approach, 19 voltage vectors are strategically selected for modulation to regulate the inverter while effectively eliminating the zero-sequence voltage (ZSV) component. Furthermore, balanced capacitor voltages are achieved by leveraging the opposite effects of redundant switching states on the neutral-point voltage. Consequently, the proposed SVM method enables simultaneous capacitor voltage balancing and elimination of the zero-sequence voltage (ZSV) component without requiring the adjustment of a balancing control factor. Simulation results are presented to validate the effectiveness of the proposed VSI topology and its modulation strategy.

2025-12-20T00:00:00+00:00 Copyright (c) 2026 Smart Systems and Devices https://jst.vn/index.php/ssad/article/view/1198 2025-12-12T08:22:55+00:00 Dung Hoang Thi Kim dung.hoangthikim@hust.edu.vn

The object of the study is to research a Helicopter Unmanned Aerial Vehicle (HUAV) made by bilateral project HNQT/SPĐP/12.19 at Hanoi University of Science and Technology. The purpose is to study the effect of the experimental set on the aerodynamic characteristics of this unmanned helicopter and study the phenomenon of aerodynamic elasticity to provide an assessment of the durability of the model in the hovering flight mode. The one-way fluid structure interaction (FSI) method which is a combination of Computational Fluid Dynamics (CFD) and Computational Structural Dynamics (CSD), has been carried out to comprehend both aerodynamic and aeroelasticity phenomena of HUAV. The CFD results show the distribution of pressure, velocity, and turbulence in accordance with the actual phenomenon. The CSD results show displacements, stress distributions, and material limit assessments. Then, a suitable operating range that meets the feasibility and possibility of flight is created. This study is a premise for further experimental studies in the process of creating a HUAV.

2025-12-20T00:00:00+00:00 Copyright (c) 2026 Smart Systems and Devices https://jst.vn/index.php/ssad/article/view/1167 2025-11-20T02:24:18+00:00 Hai Le Bui thoa.macthi@hust.edu.vn Thi-Thoa Mac thoa.macthi@hust.edu.vn

The study proposes a simple approach to design optimally a fuzzy controller in swing-up and position control for inverted pendulum-cart systems. First, the sub-fuzzy controllers to control the pendulum's swing up and the cart's position are designed separately. Each controller includes two state variables to calculate the component control forces. The combination of component control forces determines the final control force through a weight computed from a simple scheme. Parameters of sub-fuzzy controllers and those to determine the weight are optimized to minimize the system's equilibration time. The simulation results show that the proposed controller is simple to set up and optimize, has high control efficiency, is adaptable to the system's state, and is stable and robust for the system's different initial conditions and configurations. When using the proposed controller, the stabilization time of the system is reduced by 14.5%, the maximum control force is reduced by 32.6%, and the pendulum length is increased by 50% compared to fuzzy controllers in the published studies. The approach of the present work can be applied to control various underactuated systems as well as in the motion control of mobile robot models.

2025-11-20T00:00:00+00:00 Copyright (c) 2025 https://jst.vn/index.php/ssad/article/view/1199 2025-12-12T08:41:38+00:00 Bao Duy Thanh Tran nhung.lethituyet@hust.edu.vn Duc Huy Ta huy.td@gcool.com.vn Dinh Quan Tran quantd@gcool.com.vn Dinh Quy Vu quy.vudinh@hust.edu.vn Thi Tuyet Nhung Le nhung.lethituyet@hust.edu.vn

The study presents the design and optimization process for high-performance ceiling fans aimed at energy savings and environmental protection. Utilizing Computational Fluid Dynamics (CFD) numerical simulations with the k-ω SST turbulence model and the multi-reference frame (MRF) method, various fan blade configurations were evaluated to enhance airflow and improve energy consumption efficiency. The design process focused on selecting the aerodynamic profile (airfoil), optimizing the twist angle, and distributing the chord length. Among the tested options, the optimized version yielded superior results in airflow distribution, torque, and noise characteristics. The grid convergence analysis with 11.5 million elements validated the accuracy of the simulations. The addition of winglets helped reduce tip vortex phenomena and sound intensity, lowering it by up to 20 dB. A frequency spectrum analysis model using the Fast Fourier Transform (FFT) was applied to assess the sound characteristics in detail and identify noise-causing frequency components. The final design meets technical requirements and manufacturability, achieving optimal performance at a speed range of 220–225 rpm for the 5-blade configuration and 265 rpm for the 3-blade configuration. Energy performance parameters were measured according to Vietnam national standards (TCVN) for electric fans, validating the simulation results and ensuring suitability under real operating conditions.

2025-12-20T00:00:00+00:00 Copyright (c) 2026 Smart Systems and Devices https://jst.vn/index.php/ssad/article/view/1001 2025-11-03T09:11:19+00:00 Mr Thanh Loi Dang dangloi363@gmail.com Dr. Computer Science Kim Ngan Nguyen ngannguyen@neu.edu.vn Assoc. Prof Nhat Quang Dinh quang.dinh@tlu.edu.vn Assoc. Prof Ngoc Doanh Nguyen doanh.nn@vinuni.edu.vn

Climate change exacerbates the frequency and intensity of flood events, presenting substantial threats to human lives and infrastructure. Consequently, accurate and timely water level forecasting systems are critical for effective early warning dissemination and rapid disaster response. While numerous studies in Vietnam have focused on river water level prediction, a notable gap exists in the specific area of continuous, multi-hour time series forecasting. This study addresses this gap by proposing a novel modeling approach for forecasting future water levels at the Le Thuy station on the Kien Giang river in Quang Tri province. The proposed models leverage historical hydrological observations from multiple upstream stations to predict water level sequences at the Le Thuy station over continuous horizons of 6, 12, and 24 hours. The methodology employs advanced deep learning techniques, specifically Autoencoder, Long Short-Term Memory (LSTM) networks, and Attention mechanism, with each forecast horizon being modeled independently. Experimental results demonstrate the models’ robust capability to accurately capture both rising and falling water level trends. The forecasted sequences exhibit strong alignment with observed values, even during periods of rapid fluctuation. Point-wise prediction errors are consistently low, indicating high forecasting precision. Crucially, the models maintain their effectiveness during extreme floodevents,successfully predicting both the magnitude and timing of floodpeaks.

2025-11-20T00:00:00+00:00 Copyright (c) 2025 Smart Systems and Devices https://jst.vn/index.php/ssad/article/view/1259 2026-01-19T09:11:22+00:00 Thi Minh Chau Le dmquan@dut.udn.vn Duc Tung Le dmquan@dut.udn.vn Hong Duy An Nguyen dmquan@dut.udn.vn Minh Quan Duong dmquan@dut.udn.vn

With the rapid proliferation of renewable energy sources (RES) in modern power grids, the application of operational optimization strategies has become pivotal in maintaining system efficiency and stability. In particular, the hybrid deep learning model long short-term memory (LSTM) neural network combined with inductive conformal prediction (ICP) significantly enhances the accuracy of renewable energy forecasting and management. This paper presents the integration of big data analytics (BDA) with the LSTM+ICP framework for optimizing smart grid operations, particularly under real-time conditions. A suite of machine learning and deep learning models, especially the hybrid LSTM+ICP, is deployed to address critical challenges such as renewable output forecasting, load balancing, and fault detection. Owing to its capability to capture temporal dependencies and generate reliable prediction intervals, the LSTM+ICP model achieved a mean absolute percentage error (MAPE) of 2.91%, thus improving the reliability of renewable energy scheduling and enabling more efficient resource allocation. The implementation of real-time BDA in conjunction with LSTM+ICP reduced load variance by 44.3%, peak demand by 24.2%, and frequency deviations by 52.9%, thereby strengthening grid reliability and operational stability. In predictive maintenance, the LSTM+ICP model achieved a detection accuracy of 97.2% with an average lead time of 6.2 hours, enabling proactive interventions and minimizing fault risks. For system optimization, the application of reinforcement learning augmented by BDA led to a 33.9% reduction in power losses, a 22.4% increase in voltage stability, and a 29.1% decrease in reactive power, thereby enhancing operational efficiency. From both economic and environmental perspectives, the BDA-driven approach resulted in monthly cost savings of C36,150 and a 30.2% reduction in CO2 emissions, demonstrating the efficacy and sustainability of the proposed methodology.

Copyright (c) 2026