• IEMEK Journal of Embedded Systems and Applications
• /
• v.14 no.5
• /
• pp.227-237
• /
• 2019

Ultra-low latency control is one of the characteristics of 5G cellular network services, which means that the control loop is handled in milliseconds. To achieve this, it is necessary to identify time delay factors that occur in all components related to CPS control loop, including new 5G cellular network elements such as MEC, and to optimize CPS control loop in real time. In this paper, a novel CPS architecture for ultra-low latency control of CPS is designed. We first define the ultra-low latency characteristics of CPS and the CPS concept model, and then propose the design of the control loop performance monitor (CLPM) to manage the timing information of CPS control loop. Finally, a case study of MEC-based implementation of ultra-low latency CPS reviews the feasibility of future applications.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.07a
• /
• pp.539-540
• /
• 2023

최근 물체 인식 모델의 성능을 개선하기 위한 다양한 연구가 진행 중이다. 본 논문에서는 K-means 기반 앵커박스 선정 기법을 적용한 새로운 물체 인식 모델 성능 개선 방법을 제안한다. 제안된 방법은 항만 내 설치된 컨테이너 사고를 예방하기 위한 컨테이너 사고위험도 분류 모델에 적용하여 성능 평가를 하였다. 특히, 컨테이너 사고위험도 분류 모델은 작은 물체를 인식해야 하며 이런 환경에서는 기존 물체 인식 모델 성능이 낮게 나타난다. 본 논문에서는 제안한 K-means 기반 앵커박스 선정 기법을 적용하여 물체 인식 모델 성능이 개선됨을 확인하였디.

• ETRI Journal
• /
• v.45 no.4
• /
• pp.666-677
• /
• 2023

In the last few years, convolutional neural networks (CNNs) have demonstrated good performance while solving various computer vision problems. However, since CNNs exhibit high computational complexity, signal processing is performed on the server side. To reduce the computational complexity of CNNs for edge computing, a lightweight algorithm, such as a MobileNet, is proposed. Although MobileNet is lighter than other CNN models, it commonly achieves lower classification accuracy. Hence, to find a balance between complexity and accuracy, additional hyperparameters for adjusting the size of the model have recently been proposed. However, significantly increasing the number of parameters makes models dense and unsuitable for devices with limited computational resources. In this study, we propose a novel MobileNet architecture, in which the number of parameters is adaptively increased according to the importance of feature maps. We show that our proposed network achieves better classification accuracy with fewer parameters than the conventional MobileNet.

• Geophysics and Geophysical Exploration
• /
• v.10 no.4
• /
• pp.275-284
• /
• 2007

For precise interpretation of magnetotelluric (MT) data distorted by irregular surface terrain, topography effects are investigated by computing apparent resistivities, phases, tippers and induction vectors for a three-dimensional (3D) hill-and-valley model. To compute MT responses for the 3D surface topography model, we use a 3D MT modeling algorithm based on an edge finite-element method which is free from vector parasites. Distortions on the apparent resistivity and phase are mainly caused by distorted currents that flow along surface topography. The distribution of tipper amplitudes over both hill and valley are the same, while the tipper points toward the center of hill and the base of the valley. The real part of induction vector also points in the same direction as that of tipper, while the imaginary part in the opposite direction.

• Journal of Korea Multimedia Society
• /
• v.20 no.8
• /
• pp.1369-1378
• /
• 2017

Pub/Sub (Publish/Subscribe) paradigm is a simple and easy to use model for interconnecting applications in a distributed environment. In general, subscribers register their interests in a topic or a pattern of events and then asynchronously receive events matching their interest, regardless of the events' publisher. In order to build a low latency lightweight pub/sub system for Internet of Things (IoT) services, we propose a GQSFPS (Group Quorum System-based Fog Pub/Sub) system that is a core component in the event-driven service oriented architecture framework for IoT services. The GQSFPS organizes multiple installed pub/sub brokers in the fog servers into a group quorum based P2P (peer-to-peer) topology for the efficient searching and the low latency accessing of events. Therefore, the events of IoT are cached on the basis of group quorum, and the delay-sensitive IoT applications of edge devices can effectively access the cached events from group quorum fog servers in low latency. The performance of the proposed GQSFPS is evaluated through an analytical model, and is compared to the GQPS (grid quorum-based pud/sub system).

• Journal of computational fluids engineering
• /
• v.4 no.1
• /
• pp.53-61
• /
• 1999

Two-dimensional, unsteady, incompressible and compressible Navier-Stokes codes are developed for the computation of the viscous turbulent flow over high-lift airfoils. The compressible code involves a conventional upwind-differenced scheme for the convective terms and LU-SGS scheme for temporal integration. The incompressible code with pseudo-compressibility method also adopts the same schemes as the compressible code. Three two-equation turbulence models are evaluated by computing the flow over single and multi-element airfoils. The compressible and incompressible codes are validated by predicting the flow around the RAE 2822 transonic airfoil and the NACA 4412 airfoil, respectively. In addition, both the incompressible and compressible code are used to compute the flow over the NLR 7301 airfoil with flap to study the compressible effect near the high-loaded leading edge. The grid systems are efficiently generated using Chimera overlapping grid scheme. Overall, the κ-ω SST model shows closer agreement with experiment results, especially in the prediction of adverse pressure gradient region on the suction surfaces of high-lift airfoils.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.32 no.3
• /
• pp.487-499
• /
• 2022

Machine learning such as deep learning have been widely used in recent years. Recently deep learning is performed in a trusted execution environment such as ARM TrustZone to improve security in edge devices and embedded devices with low computing resource. To mitigate this problem, we propose TPMP that efficiently uses the limited memory of TEE through DNN model partitioning. TPMP achieves high confidentiality of DNN by performing DNN models that could not be run with existing memory scheduling methods in TEE through optimized memory scheduling. TPMP required a similar amount of computational resources to previous methodologies.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.12
• /
• pp.3904-3922
• /
• 2022

As a research hotspot, pedestrian detection has a wide range of applications in the field of computer vision in recent years. However, current pedestrian detection methods have problems such as insufficient detection accuracy and large models that are not suitable for large-scale deployment. In view of these problems mentioned above, a lightweight pedestrian detection and early warning method using a new model called you only look once (Yolov5) is proposed in this paper, which utilizing advantages of Yolov5s model to achieve accurate and fast pedestrian recognition. In addition, this paper also optimizes the loss function of the batch normalization (BN) layer. After sparsification, pruning and fine-tuning, got a lot of optimization, the size of the model on the edge of the computing power is lower equipment can be deployed. Finally, from the experimental data presented in this paper, under the training of the road pedestrian dataset that we collected and processed independently, the Yolov5s model has certain advantages in terms of precision and other indicators compared with traditional single shot multiBox detector (SSD) model and fast region-convolutional neural network (Fast R-CNN) model. After pruning and lightweight, the size of training model is greatly reduced without a significant reduction in accuracy, and the final precision reaches 87%, while the model size is reduced to 7,723 KB.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.19 no.3
• /
• pp.151-158
• /
• 2024

Although on-device artificial intelligence (AI) has gained attention to diagnosing machine faults in real time, most previous studies did not consider the model retraining and redeployment processes that must be performed in real-world industrial environments. Our study addresses this challenge by proposing an on-device AI-based real-time machine fault diagnosis system that utilizes continual learning. Our proposed system includes a lightweight convolutional neural network (CNN) model, a continual learning algorithm, and a real-time monitoring service. First, we developed a lightweight 1D CNN model to reduce the cost of model deployment and enable real-time inference on the target edge device with limited computing resources. We then compared the performance of five continual learning algorithms with three public bearing fault datasets and selected the most effective algorithm for our system. Finally, we implemented a real-time monitoring service using an open-source data visualization framework. In the performance comparison results between continual learning algorithms, we found that the replay-based algorithms outperformed the regularization-based algorithms, and the experience replay (ER) algorithm had the best diagnostic accuracy. We further tuned the number and length of data samples used for a memory buffer of the ER algorithm to maximize its performance. We confirmed that the performance of the ER algorithm becomes higher when a longer data length is used. Consequently, the proposed system showed an accuracy of 98.7%, while only 16.5% of the previous data was stored in memory buffer. Our lightweight CNN model was also able to diagnose a fault type of one data sample within 3.76 ms on the Raspberry Pi 4B device.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.10
• /
• pp.3438-3457
• /
• 2022

As location-based services (LBS) are widely used in vehicular ad-hoc networks (VANETs), location privacy has become an utmost concern. Spatial cloaking is a popular location privacy protection approach, which uses a cloaking area containing k-1 collaborative vehicles (CVs) to replace the real location of the requested vehicle (RV). However, all CVs are assumed as honest in k-anonymity, and thus giving opportunities for dishonest CVs to submit false location information during the cloaking area construction. Attackers could exploit dishonest CVs' false location information to speculate the real location of RV. To suppress this threat, an edge-assisted Trusted Collaborative Anonymity construction scheme called TCA is proposed with trust mechanism. From the design idea of trusted observations within variable radius r, the trust value is not only utilized to select honest CVs to construct a cloaking area by restricting r's search range but also used to verify false location information from dishonest CVs. In order to obtain the variable radius r of searching CVs, a multiple linear regression model is established based on the privacy level and service quality of RV. By using the above approaches, the trust relationship among vehicles can be predicted, and the most suitable CVs can be selected according to RV's preference, so as to construct the trusted cloaking area. Moreover, to deal with the massive trust value calculation brought by large quantities of LBS requests, edge computing is employed during the trust evaluation. The performance analysis indicates that the malicious response of TCA is only 22% of the collaborative anonymity construction scheme without trust mechanism, and the location privacy leakage is about 32% of the traditional Enhanced Location Privacy Preserving (ELPP) scheme.