Mobile QR Code QR CODE

Multi-dimensional Task Offloading using Deep Learning for Vehicular Edge Computing Networks: A Survey

(S. Syed Abuthahir) ; (J. Selvin Paul Peter)

Vehicular networks must support connection ubiquity and high levels of services for a large number of vehicles. In vehicular networks, mobile edge computing (MEC) is considered a viable technique, utilizing computing resources at the edge of wireless access networks. This survey examines effectual task offloading systems forvehicular edge computing (VEC) networks. Several deep learning methods have been suggested to enable task offloading schemes. Therefore, task offloading optimality is a main research topic in fog computing. The main aim of this work is to offer readers an overview of the journey from the task offloading concept to its mathematic problem formulation. This survey introduces fog computing and the process of task offloading that emulates numerous aspects of optimization. Subsequently, a number of machine learning and deep learning methodologies are employed in task offloading,and the challenge of fog computing is discussed. This paper provides a detailed statistical analysis from papers published between 2019 and 2022.

Deep Neural Network Entrepreneurial Project Recommendation Model for the Integration of Industry, Education, and Entrepreneurship Needs of Students

(Tao Long) ; (ZeKun Wang)

As the size of the entrepreneurship project information platform grows, it is becoming increasingly difficult for student users to find in-demand entrepreneurship projects that integrate industry and education comprehensively and rapidly. The severe information overload leads to poor accuracy of recommendation results. This study addressed these problems based on Deep Neural Networks (DNNs) and Matrix Decomposition Algorithms (MDAs) by combining a Convolutional Neural Network (CNN), word embedding, and one-hot coding techniques. The DNN-MF model was used to extract the entrepreneurial needs and implicit features of students. The DNN-MF model designed for the study was also improved and incorporated with student user features, i.e., the DNN-DNN2 model was constructed. The experiments showed that the Root Mean Square Error (RMSE) of the DNN-MF model was lower than that of the Convolution Matrix Factorization (ConvMF) and Probabilistic Matrix Factorization (PMF) by 0.1190 and 0.1677, respectively. The RMSE of the DNN-DNN2 model was lower than that of the DNN-MF model, and the recommendation accuracy of the study model was 2.35% higher than that of the DNN-DNN1 model, which did not incorporate the student user characteristics. These results showed that the proposed recommendation model for entrepreneurial projects was significantly better than the current popular ones. Moreover, the model could complete the task of recommending entrepreneurial projects faster and more accurately, effectively solving the cold start problem of users and projects, which has certain practical significance.

SMAGNet: Scaled Mask Attention Guided Network for Vision-based Gait Analysis in Multi-person Environments

(Hosang Yu) ; (Jaechan Park) ; (Kyunghun Kang,) ; (Sungmoon Jeong)

Clinical gait analysis plays a key role in diagnosing and managing neurodegenerative diseases such as Parkinson’s disease. In recent years, vision-based gait analysis methods have emerged as promising non-invasive approaches to quantify gait characteristics. However, most methods assume single-person situations, but multi-person situations are more common in real-world medical settings. In this paper, we propose a novel mask-guided attention model called a Scaled Mask Guided Attention Network (SMAGNet), which exploits a target person's detection result to address multi-person issues. SMAGNet utilizes a detection box as a mask label to predict attention maps that highlight patients’ gait features and progressively refines the maps for accurate analysis. Experimental results show that the mean absolute percentage error (MAPE) was improved by up to 20% for the target spatio-temporal gait variable compared to the baseline 3D CNN (Convolutional Neural Networks). Moreover, we achieved significantly better performance compared to other methods, including a recent state-of-the-art gait recognition model named GaitBase. These results showcase SMAGNet’s effectiveness in multi-person gait analysis and its potential for real-world clinical use.

Efficient Task Offloading in Vehicular Fog Networks

(Ihsan Ullah) ; (Byung-Seo Kim)

Vehicular fog computing (VFC) has emerged as the key enabler in reducing computation processing delays and relieving base station (fog node) workload by utilizing the underutilized resources of vehicles during peak traffic hours. The base station distributes computation tasks among the available fog nodes. In VFC, efficient task placement and judicious resource utilization that can provide timely computations to delay-intolerant applications within the strict latency budget are highly challenging. Therefore, this paper proposes an efficient resource utilization and task placement strategy in VFC to reduce the computation latency and network resource utilization. The fog node adopts the weighted max-min fair share (WMMFS) to prioritize the received computation tasks and distribute the tasks among the available vehicles based on tasks’ priority. An efficient strategy is devised where the high-priority tasks are offloaded toward the most reliable vehicles to uphold the QoS and QoE application. Extensive simulations showed that the proposed offloading scheme outperformed the random offloading scheme.

Detection and Prevention of Black Hole Attack using Tree Hierarchical Deep Convolutional Neural Network and Enhanced Identity-based Encryption in Vehicular Ad Hoc Network

(K. Lakshmi Narayanan) ; (R. Naresh)

A method for the recognition and prevention of a black hole attack is proposed using a tree hierarchical deep convolutional neural network (THDCNN)and enhanced identity based encryption in a vehicular ad hoc network (VANET). Automobiles are organized with a cluster formula using an improved k-means clustering algorithm. Following the cluster formation, cluster head (CH) selection is done using a balancing composite motion optimization (BCMO) algorithm. After selecting the cluster head, the entrance of a spiteful node occurs in the cluster. The THDCNN is proposed for classifying a cluster node as i) a black hole attack node or ii) a normal node. If a black hole attack node is found, the attack node information is communicated to an individual CH, which makes a final decision. Otherwise, the standard node data is encrypted with the enhanced identity-based encryption algorithm. Finally, the proposed method prevents the normal node data from the attacker. Thus, the proposed method attains higher accuracy and lower computational time than other methods.

The Application of Mobile Technology in the Innovation of the Internationalization Model of English Teaching

(Bin Li)

The “explosive” development of information technology and the acceleration of teaching reform have made it difficult to adapt the traditional “cramming” English teaching method to the development of education and affect the improvement of students’ comprehensive quality and the international innovation of teaching mode. Therefore, based on the current situation and characteristics of English teaching, the research proposes that the dynamic prediction of teaching information will be realized with the help of an improved radial basis function neural network algorithm based on mobile technology, i.e., the support and confidence in association rules will be used as the basis for RBF algorithm to select the data center, and a mobile intelligent teaching system will be built to promote the transformation of teaching mode. The results showed that the overall sample output error of the improved RBF algorithm was less than 3%. The generalization performance was good. The classification accuracy of teaching data was more than 93%, and the teaching correlation was basically more than 0.15. The system platform can effectively extract the characteristic factors of teaching importance, which makes the score of teachers’ and students’ dissatisfaction with the effect of English teaching decrease by more than 55%. The above results showed that the teaching system supported by mobile technology can effectively realize the dynamic management of English teaching activities and provide some technical and theoretical support for teachers to adjust their teaching plans.

Neuro-facial Fusion for Emotion AI: Improved Federated Learning GAN for Collaborative Multimodal Emotion Recognition

(Recognition D. Saisanthiya) ; (P. Supraja)

In the context of artificial intelligence technology, an emotion recognition (ER) has numerous roles in human lives. On the other hand, the emotion recognition techniques most currently used perform poorly in recognizing emotions, which limits their wide spread use in practical applications. A Collaborative Multimodal Emotion Recognition through Improved Federated Learning Generative Adversarial Network (MER-IFLGAN) for facial expressions and electro encephalogram (EEG) signals was proposed to reduce this issue. Multi-resolution binarized image feature extraction (MBIFE) was initially used for facial expression feature extraction. The EEG features were extracted using the Dwarf Mongoose Optimization (DMO) algorithm. Finally, IFLGAN completes the Emotion recognition task. The proposed technique was simulated in MATLAB. The proposed technique achieved 25.45% and 19.71% higher accuracy and a 32.01% and 39.11% shorter average processing time compared to the existing models, like EEG based Cross-subject and Cross-modal Model (CSCM) for Multimodal Emotion Recognition (MER-CSCM) and Long-Short Term Memory Model (LSTM) for EEG Emotion Recognition (MER-LSTM), respectively. The experimental results of the proposed model shows that complementing EEG signals with the features of facial expression could identify four types of emotions: happy, sad, fear, and neutral. Further more, the IFLGAN classifier can enhance the capacity of multimodal emotion recognition.

A Comprehensive Review of Cyber-physical System (CPS)-based Approaches to Robot Services

(Soomin Shim) ; (Joon Young Kim) ; (Seong Wook Hwang) ; (Jeong Min Oh) ; (Bo Kyung Kim) ; (Ju Hee Park) ; (Dong Jin Hyun,) ; (Heesang Lee)

Robotics is one of the leading technologies that can support and assist human activities and physical tasks. Recent advancements regarding the metaverse and virtual reality have allowed robots to operate through teleoperation or in a remote virtual space in real time. In addition to these technical advancements, various service and business items are being developed and deployed throughout virtual spaces and in robotics. In this context, it is necessary to approach virtual space and robotics research based on a cyber-physical system (CPS). However, to identify state-of-the-art services and items associated with CPS-based robotics research, we need in-depth analyses and follow-up of current research and development trends in virtual space and in robotics from academic and industrial perspectives. The present work investigates current trends in, and considerations of, robotics in virtual space by taking comprehensive approaches, specifically, the systematic literature review (SLR) and technical standards analysis (TSA). In the SLR, we finalized a selection of 36 publications, and then proceeded with trend and technical-item analyses. In the TSA, we selected four standards documents and analyzed the contents and critical aspects of each standard. The analysis results are interconnected to provide comprehensive insights into, and considerations for, CPS-based robot services.

Predictive Analytics for Municipal Public Security: A Comprehensive Scheme Integrating Big Data and Regression Models

(Models Moonsik Kang) ; (Yonggyu Jung)

As the operation of the municipal police system has started, people's attention has been focused on whether the demand for security personnel is appropriate. Municipal police is an executive organization of local government with autonomous administrative authority and autonomously provides municipal police service for residents in the jurisdiction. In other words, it is an organization that conducts police activities under the responsibility of the local government. This study designed a self-policing demand prediction system using big data analysis and a regression model based on regional characteristics, and the performance was evaluated by analyzing the correlations. The model proposed in this paper is a big data analysis model that extracts major variables using correlation analysis between the collected variables and the current demand for municipal police personnel. The performance of the proposed scheme was evaluated by generating and analyzing the appropriate number of people in need of security in a specific G area. The results confirmed that the prediction was accurate.

Approach to Smart Mobility Intelligent Traffic Signal System based on Distributed Deep Reinforcement Learning

(Young-Sik Lee)

Smart mobility intelligent traffic services have become critical in intelligent transportation systems (ITS). This involves using advanced sensors and controllers and the ability to respond to real-time traffic situations at intersections, alleviate congestion, and generate policies to prevent traffic jams. Deep reinforcement learning (DRL) provides a natural framework for processing tasks. In DRL, each intersection can control itself and coordinate with neighbors to achieve optimal network-wide policies. On the other hand, comparing approaches remains a challenging task due to the existence of numerous possible configurations. This research performs a critical comparison of various traffic controllers in the literature. Hence, using a nonlinear approximator for coordination mechanisms and enhancing observability at each intersection are key performance drivers.

Enhanced Multi-trilateration Method to Mitigate Shadow Fading for ROA-based Indoor Sensor Localization

(Kyunghee Seo) ; (Hoki Baek)

Sensor localization is difficult in indoor environments because of the multipath fading and shadow fading caused by obstacles [1]. Multipath fading and shadow fading cause large localization errors that make the sensor node fail in the sensor localization process. This paper highlights the indoor sensor localization concept and proposes a ROA (Received signal strength Of Arrival)-based sensor localization scheme. The proposed scheme called RERR (Range Expansion and Range Reduction) can mitigate the above-mentioned problems for indoor environments. The RERR-based sensor nodes virtually expand or reduce the measured range to succeed in localization regardless of the effects of multipath fading and shadow fading over communicating sensor nodes. The simulation results verified the effectiveness of the proposed scheme, showing small RMS errors and a large number of sensor nodes succeeding in the localization process.