https://jst.vn/index.php/ssad en-US Mon, 15 Sep 2025 00:00:00 +0700 OJS 3.2.1.1 http://blogs.law.harvard.edu/tech/rss 60 https://jst.vn/index.php/ssad/article/view/1029 The overcrowding of patients in hospitals has become a significant problem in many countries. To address this problem, the application of information technology, Internet of Things, and artificial intelligence is increasingly being integrated into major hospitals worldwide to support doctors and nurses in patient care and improve the quality of healthcare services. The study proposes an automated medication dispensing device designed to help nurses administer medication to patients. The device is built around a Raspberry Pi 4 embedded computer and utilizes face recognition technology and QR code scanning to verify patient identities, ensuring that medication is dispensed to the correct person. In addition, QR codes are used to authenticate information in medication packages, ensuring that the correct medication is provided to the patient. The device also has the ability to replenish medication supplies and notify nurses in cases where incorrect medication is administered. Thanh Phuc Chu, Minh Huy Vu, Quoc Anh Do, Vinh Tran-Quang Copyright (c) 2025 https://creativecommons.org/licenses/by-nc/4.0 https://jst.vn/index.php/ssad/article/view/1029 https://jst.vn/index.php/ssad/article/view/1030 In this work, a multi-task learning model for age, gender, and emotion recognition on edge processing is developed. The multi-task model is based on the backbone of MobileNetV2 in which the three last layers are customized to have three outputs for age, gender, and emotion. The model was trained and tested on a dataset which is the combination of the well-known dataset, namely Internet Movie Database (IMDB) and our self-collected dataset. The trained model is then optimized and quantized to be implemented on Neural Processing Unit (NPU) of the chip RK3588 from Rockchip on Orange PI plus hardware platform. Experimental evaluation on several testcase was performed. It is known that the multi-task model outputs prediction accuracy as high as single-task model while significantly reducing computational processing requirements. On Orange PI platform, the highest prediction accuracy for age, gender and emotion are 3.485 MAE, 98.281%, and 93.917%, respectively. The computational performance reaches 285.7 frames per second as the highest. These results have a high potential for many practical applications on edge devices. Ha Xuan Nguyen, An Dao, Duc Quang Tran, Minh Tuan Dang Copyright (c) 2025 https://creativecommons.org/licenses/by-nc/4.0 https://jst.vn/index.php/ssad/article/view/1030 https://jst.vn/index.php/ssad/article/view/1031 Population aging is considered one of the prominent trends of the 21st century, manifested by the rapidly increasing proportion of elderly people, which profoundly impacts all aspects of a country, territory, and even globally. As the human body ages and its functions decline, falls become a significant risk to both physical and mental health, causing severe injuries to the elderly and creating a financial burden for their families. Therefore, proactive protective devices are highly valued for development to protect the elderly's bodies from serious injuries in the event of a fall. Airbags, which are known as safety components in automobiles, are designed to deploy during a collision to protect the occupants by reducing the force applied to the body by the dashboard, steering wheel, or any part of an automobile. Smart airbags are also designed to protect the elderly in a fall. The airbag will be inflated rapidly before falling to hug the body area that needs protection as a cushion to reduce the impact force between the body and the ground, thereby reducing injury to the body when falling. This paper provides an overview of research on smart airbags for protecting elderly people when falling, the sensors used in smart airbags, and the parameters that determine the moment of a fall. Thanh Thao Phan, Xuan Hien Vu, Thi Mai Anh Le Copyright (c) 2025 https://creativecommons.org/licenses/by-nc/4.0 https://jst.vn/index.php/ssad/article/view/1031 https://jst.vn/index.php/ssad/article/view/1032 The State of Charge (SOC) of the lithium-ion battery plays a vital role in monitoring and optimizing the performance of the battery management system (BMS). Traditional Kalman filter (KF) algorithm requires an accurate understanding of the dynamic model of the system and usually contains unknown statistical noises, which can make the SOC estimation inaccurate. To overcome the problem, this paper proposes a modified Kalman filter, namely Kalman-LSTM, which integrates the Long Short-Term Memory (LSTM) into the KF framework. By incorporating a neural network, the method preserves the data efficiency and interpretability of traditional algorithms while simultaneously learning the dynamic behavior of the system. The accuracy of the Kalman-LSTM method is tested using four datasets: DST, BJDST, FUDS, and US06. The SOC estimation results are then compared with different KF variants, including EKF, UKF, and AKF. The experimental results demonstrate that the proposed model has superior accuracy compared to benchmark models across various working conditions. Hoang Minh Giang Nguyen, Trong Thanh Nguyen, Thi Hoai Thu Nguyen, Hirotaka Takano, Duc Tuyen Nguyen Copyright (c) 2025 https://creativecommons.org/licenses/by-nc/4.0 https://jst.vn/index.php/ssad/article/view/1032 https://jst.vn/index.php/ssad/article/view/1036 The article focuses on developing a predictive and optimization model for surface roughness in ultra-precision turning (UPT) using a diamond cutter on the spherical surface of Al6061 material. An experimental model with 30 experiments was established, considering three input parameters: spindle speed, feed rate, and depth of cut. The results from this model were collected to create an input dataset for an artificial neural network (ANN) to build a surface roughness prediction model. The ANN structure 3-5-10-1 provided the best prediction results, with Coefficient of Determination ( ) is 0.98, Mean Absolute Percent Error ( ) is 13.36%, Mean Square Error ( ) is 0.68, and Root Mean Square Error ( ) is 0.82. Additionally, the artificial bee colony (ABC) algorithm was employed to determine the optimal cutting parameters that minimize surface roughness. The results indicated that the minimum roughness value achieved was 0.76 nm with the cutting parameters: spindle speed of 823 rev/min, feed rate of 13 mm/min, and depth of cut of 1 µm. Moreover, the effects of different cutting parameter combinations on surface roughness were analyzed and evaluated. The integration of the ANN model with the ABC algorithm enables a reliable prediction model for surface roughness and demonstrates high efficiency in optimizing the objective function. This research contributes valuable insights into surface roughness prediction and optimization in ultra-precision turning of Al6061 material. Furthermore, the proposed modeling and optimization approach can be extended to other materials and the processing of aspherical and diffractive surfaces. Nghia Duc Hoang, Xuan Bien Duong, Kim Hoa Bui, Kim Hung Nguyen, Manh Tung Do, Viet Hung Ngo, Khanh Nghia Truong Copyright (c) 2025 https://creativecommons.org/licenses/by-nc/4.0 https://jst.vn/index.php/ssad/article/view/1036 https://jst.vn/index.php/ssad/article/view/1037 The paper studies the effect of some shape parameterization techniques on automatic two-dimensional aerodynamic shape optimization using the discrete adjoint method. In this paper, the Hicks-Henne Bump Functions (HHBF) technique and the Free-form Deformation (FFD) control points technique are used to parameterize the shape of the NACA 0012 airfoil. First, this paper makes a full, detailed description of the shape optimization workflow, including Euler equations, geometry parameterization techniques, discrete adjoint method, gradient evaluation, optimization algorithm, and mesh deformation. Second, it explores how shape parameterization techniques are implemented in the optimization problem. Finally, the results are evaluated to compare the efficiency of the mentioned techniques. The results suggest that, in general, both techniques were shown to be equally effective as geometry parameterization methods for the shape optimization problem. However, it appears that the HHBF technique demonstrates better performance with fewer design iterations compared to that of FFD technique. On the other hand, FFD shows stability and a smoother decrease in drag values, while HHBF exhibits greater unsteadiness during the optimization process. Trung-Huy Nguyen, Gia-Long Hoang, Huy-Duc Nguyen, Quoc-Bao Nguyen, Van-Sang Pham Copyright (c) 2025 https://creativecommons.org/licenses/by-nc/4.0 https://jst.vn/index.php/ssad/article/view/1037 https://jst.vn/index.php/ssad/article/view/1038 These days, collaborative robots (Cobot) are well-known for their adaptability and several uses. With a 7-degree-of- freedom cobot model, the work emphasises more the flexibility as a benefit to overcome the joint constraints usually faced in robotic systems. Applying the Jacobian matrix approach with its null-space helps one to consider the kinematics and dynamics issue in order to prevent joint limitation. Whereas orientation is stated using Roll-Pitch-Yaw angles, the inverse kinematics problem involves parameterising the robot's end-effector by its position in Cartesian coordinates. By means of the null space of the Jacobian matrix, one can escape joint constraints in robotic mobility. Emphasising great accuracy in computations with two path planning: rectilinear and curved path, this research offers a thorough study of the trajectory tracking control dynamics. Different modules and simulation results of various challenges have been numerically implemented in MATLAB-Simulink and the ROS environment to show the efficiency of the suggested method after analysis of the computations. Phuong Thao Thai, Quang Hoang Nguyen, Van Phong Dinh Copyright (c) 2025 https://creativecommons.org/licenses/by-nc/4.0 https://jst.vn/index.php/ssad/article/view/1038 https://jst.vn/index.php/ssad/article/view/1039 The aim of the study was to estimate aerodynamic characteristics of Unmanned Aerial Vehicle (UAV) in flight mode at a flight altitude of 100m using Computational Fluid Dynamics (CFD) in ANSYS software. Firstly, aerodynamic characteristics of Unmanned Aerial Vehicles (UAV) was estimated for steady case and then for unsteady case to investigate effect of time in comparison the different of air flow for these cases. Secondly, effect of initial velocity and frequency of fluctuation were performed using User Define Function (UDF) in ANSYS software. The motion speed was changed in a sinusoidal form with amplitudes of 10, 15, 20, and 25 m/s and two oscillation frequencies of 0.1 and 0.2 Hz. Research results showed that speed amplitude did not affect the aerodynamic quality of the aircraft in the selected speed range. Meanwhile, the frequency of oscillation affected quite a lot on the aerodynamics of the aircraft. Finally, at velocity 20 m/s, the side slip angle was varied from 0 to 12 degrees to carry out effect of slide slip angle. The UAV tended to lose lift, increase drag, leading to a strong reduction of aerodynamic quality when increasing side slip angle. This indicates an unfavourable influence of side slip angle on the aerodynamic characteristics of the UAV. Thi Thai Le, Thi Kim Dung Hoang, Thi Huyen Thuong Ho Copyright (c) 2025 https://creativecommons.org/licenses/by-nc/4.0 https://jst.vn/index.php/ssad/article/view/1039