Green, Pervasive, and Cloud Computing

Cloud computing offloads user tasks to remote cloud servers, which can effectively enhance the user’s network experience, but in recent years, as the number of offloaded tasks increases and users’ real-time requirements improve, cloud services are becoming increasingly challenging to meet users’ needs. Edge computing deploys multiple edge servers around the users. The shorter distance from users can significantly reduce the transmission time of task data and avoid unpredictable network latency, which is especially suitable for normal users whose tasks are mainly data-intensive tasks. However, the large variability in the density of users in different areas and the type of computing tasks (i.e., compute-intensive and data-intensive) in the same area leads to the challenge of optimally deploying multiple edge servers. To address this challenge, we design a method named optimal placement of edge servers in the cloud environment (OPECE). First, this optimal placement problem is modeled as a constrained multi-objective optimization model with task time and server utilization as the two optimization objectives. Then, this multi-objective optimization model is optimized using the Non-dominated Sorting Differential Evolution (NSDE) algorithm. Finally, the effectiveness and superiority of OPECE are demonstrated by comparing it with the currently used methods.

In recent years, massive services that provide similar functions continue to emerge. Since services sensitive to latency and throughput are often expected to have high Quality of Service (QoS), how to accurately predict QoS has become a challenging issue. Some current deep learning (DL) based approaches usually simply concatenate the embedding vectors, without considering the correlation between embedding dimensions. Besides, the high-order feature interactions are not sufficiently learned. To this end, this paper proposes a Convolutional Neural Network based QoS prediction model with Dimensional Correlation, named QPCN. First, the two dimensional interaction features is explicitly obtained by modeling the embedding vectors. Then, the convolutional neural network is utilized to perform feature extraction and complete QoS prediction. Compared with the fully connected network, QPCN can build a deeper model and learn high-order features. In addition, the parameters of QPCN are significantly reduced, which will reduce the time and energy consumption of inference. The effectiveness of QPCN is validated by experiments on a real-world dataset.

Mobile Edge Computing (MEC) has become an indispensable way to reduce the execution delay of devices. However, for some devices located far away from the MEC server, the transmission delay of communication with MEC is still large. In this case, we consider using multiple relay devices to assist Internet of Things (IoT) devices to communicate with MEC servers. To enhance the energy efficiency of the system, we introduce Energy Harvesting (EH) devices to provide energy for the IoT devices. Our objective is to maximize the utilization of EH devices while minimizing the overall delay in task offloading for the IoT devices. We tackle the problem by formulating it as a Markov Decision Problem (MDP). However, due to the significant expansion of the state space, traditional methods such as relative value iteration and linear iterative reconstruction are ineffective in solving this problem. Hence, we propose an approach called Multi-Relay Assisted Dynamic Computation Offloading (MRADCO) algorithm, which leverages the Lyapunov optimization technique. It is important to note that our proposed algorithm makes decisions solely based on the current state, without requiring the distribution information of the wireless channel and EH process. This characteristic enhances the algorithm’s practicality and reduces complexity in real-world implementations. Through rigorous theoretical derivation and comprehensive simulation experiments, we demonstrate that our algorithm is asymptotically optimal. And compared with the benchmark algorithm LODCO, our algorithm reduces the time by 50%.

The development of the software and hardware has brought about the abundance and overflow of computing resources. Many Internet companies can lease idle computing resources based on the peak and valley cycles of usage load to provide IaaS service. After the surging development, more and more companies are gradually paying attention to the specific user experience in cloud computing. But there are few works about that aspect in load balance problem. Therefore, we consider the load balancing resource allocation problem, from the perspective of user experience, mainly based on geographical distance and regional Equilibrium, combined with user usage habits, in multiple service periods. A detailed mathematical definition has been established for this problem. This article proposes a mathematical model for the problem, along with an modified CMA-ES (Covariance Matrix Adaptation Evolution Strategy) algorithm, called WS-ESS-CMA-ES algorithm, to allocate computing resources and solve the above problem. The process of using the proposed algorithm to solve the problem of cloud computing resource allocation in real scenarios is simulated, and compared with some other algorithms. The experiment results show that our algorithm performs well.

Nowadays, artificial intelligence-based tasks are imposing increasing demands on computation resources and energy consumption. Unmanned aerial vehicles (UAVs) are widely utilized in mobile edge computing (MEC) due to maneuverability and integration of MEC servers, providing computation assistance to ground terminals (GTs). The task offloading process from GTs to UAVs in UAV-enabled MEC faces challenges such as workload imbalance among UAVs due to uneven GT distribution and conflicts arising from the increasing number of GTs and limited communication resources. Additionally, the dynamic nature of communication networks and workload needs to be considered. To address these challenges, this paper proposes a Multi-Agent Deep Deterministic Policy Gradient based distributed offloading method, named DMARL, treating each GT as an independent decision-maker responsible for determining task offloading strategies and transmission power. Furthermore, a UAV-enabled MEC with Non-Orthogonal Multiple Access architecture is introduced, incorporating task computation and transmission queue models. In addition, a differential reward function that considers both system-level rewards and individual rewards for each GT is designed. Simulation experiments conducted in three different scenarios demonstrate that the proposed method exhibits superior performance in balancing latency and energy consumption.

Federated Learning (FL) has emerged as a promising approach for preserving data privacy in recommendation systems by training models locally. Recently, Graph Neural Networks (GNN) have gained popularity in recommendation tasks due to their ability to capture high-order interactions between users and items. However, privacy concerns prevent the global sharing of the entire user-item graph. To address this limitation, some methods create pseudo-interacted items or users in the graph to compensate for missing information for each client. Unfortunately, these methods introduce random noise and raise privacy concerns. In this paper, we propose FedRKG, a novel federated recommendation system, where a global knowledge graph (KG) is constructed and maintained on the server using publicly available item information, enabling higher-order user-item interactions. On the client side, a relation-aware GNN model leverages diverse KG relationships. To protect local interaction items and obscure gradients, we employ pseudo-labeling and Local Differential Privacy (LDP). Extensive experiments conducted on three real-world datasets demonstrate the competitive performance of our approach compared to centralized algorithms while ensuring privacy preservation. Moreover, FedRKG achieves an average accuracy improvement of 4% compared to existing federated learning baselines.

Addressing privacy concerns and the evolving nature of user preferences, it is crucial to explore collaborative training methods for federated recommendation models that match the performance of centralized models while preserving user privacy. Existing federated recommendation models primarily rely on static relational data, overlooking the temporal patterns that dynamically evolve over time. In domains like travel recommendations, factors such as the availability of attractions, introduction of new activities, and media coverage constantly change, influencing user preferences. To tackle these challenges, we propose a novel approach called FedNTF. It leverages an LSTM encoder to capture multidimensional temporal interactions within relational data. By incorporating tensor factorization and multilayer perceptrons, we project users and items into a latent space with time encoding, enabling the learning of nonlinear relationships among diverse latent factors. This approach not only addresses the privacy concerns by preserving the confidentiality of user data but also enables the modeling of temporal dynamics to enhance the accuracy and relevance of recommendations over time.

Now Unmanned Aerial Vehicle (UAV) with Mobile Edge Computing (MEC) severs and Device-to-Device (D2D) communications provide offload computing services for User Devices (UDs). However, the UAV has relatively high transmission latency. And D2D lacks the necessary flexibility. In this paper, we introduce a novel MEC system that utilizes the collaborative advantages of flexible movement of UAV and the low latency transmission of D2D communication to process tasks from UDs. We formulate an optimization problem focused on minimizing the tasks transmission and execution delay of UDs. The problem involves joint optimization of user scheduling, UAV trajectory, and resource allocation of Virtual Machines (VMs) on the MEC server. To tackle this non-convex problem, we propose a Deep Reinforcement Learning (DRL) algorithm with Deep Deterministic Policy Gradient (DDPG). Through simulation results, we demonstrate that DDPG reduces the latency by 41% compared to Deep Q-Network (DQN) and Actor-Critic (AC) algorithm. Our collaborative UAV-D2D model has 16% and 32% lower latency than when only the UAV or D2D works alone.

In this paper, an enhanced INGO optimization algorithm is proposed to solve the problem of large positioning error of the original DV-Hop algorithm in wireless sensor networks. By introducing cubic chaotic mapping and increasing the diversity of population initialization to expand the search scope, the sensor node location information can be collected more widely, so that the algorithm can search for the best solution as far as possible. In addition, a hybrid method of optimal - worst reverse learning and lens imaging reverse learning strategy is added to help the algorithm get rid of the local extreme value easily and improve the positioning accuracy. By comparing with the localization results of the classical DV-Hop localization algorithm, SSADV-Hop algorithm, and WOADV-Hop algorithm, the INGO algorithm reduces the average normalized localization error when the beacon node, communication radius, and total number of nodes are different.

Learning diverse and flexible locomotion strategies in uncertain environments has been a longstanding challenge for quadruped robots. Although recent progress in domain randomization has partially tackled this difficulty by training policies on a wide range of potential factors, there is still a great need for improving efficiency. In this paper, we propose a novel framework for adaptive quadruped robot locomotion learning in uncertain environments. Our method is based on data-efficient reinforcement learning and learns simulation parameters iteratively. We also propose a novel Sampling-Interval-Adaptive Identification (SIAI) strategy that uses historical parameters to optimize sampling distribution and then improve identification accuracy. Final evaluations based on multiple robotic locomotion tasks showed superiority of our method over baselines.

The increasing deployment of the Internet of Things (IoT) leads to the diversified development of IoT-based applications. However, due to the fast updates and the growing scale of IoT applications, IoT developers mainly focus on the production code but overlook the co-evolution of the corresponding test code. To facilitate the maintenance of IoT applications, this paper proposes an NLP-based approach to predict whether the test code needs to be co-changed when its production code is updated. We collected data from the most popular projects on GitHub (top 1,000 with the highest stars). Three neural encoders were employed to capture semantic features of commit messages, production code changes, and related test code. We then generated our training samples, in which the features of each sample consist of < Commit Message, Production Code Change, Test Unit Code >. Finally, a neural network model was built by learning the correlations among these features to determine the possibility of test co-evolution. We evaluated the effectiveness of our NLP-based approach on 15 widely used Python projects in the IoT domain. The evaluation result shows that the prediction accuracy of our model achieves 93%, highlighting the practical significance of our approach in the maintenance of IoT applications.

With the gradual formation of the cloud computing ecosystem, the value of cloud computing is becoming increasingly evident, and the accompanying information security issues have become core elements. The challenge lies in securely transmitting, computing, and sharing data in a cloud computing environment while keeping privacy. In this paper, we propose an access control proxy re-encryption scheme that supports inner product operations based on lattices in the standard model. First, the ciphertext is linked to the attributes, and condition for transition is associated with the re-encryption key. The proxy may re-encrypt ciphertext that meets attribute conditions instead of all ciphertext, thereby limiting the proxy’s conversion permissions. Furthermore, the user outputs the inner product value after decryption. Finally, honestly re-encryption attacks (HRA) security with increased CPA security is employed to better capture the target of proxy re-encryption. In addition, we propose a new method of hiding access policy, which uses differential privacy technology to perturb the attribute values to better safeguard the privacy of users.

The rapid development of new communication technologies such as social software and email has brought convenience, but it has also increased the distance between people. This research aims to capture micro-expressions in real-time through smart glasses and uses deep learning technology to develop a real-time emotion recognition system based on RGB color values to improve interpersonal communication. We trained a multi-layer fully deep neural network (DNN) model using the CASME2 dataset to effectively learn the association between facial expressions and emotions. The experimental results of model show that the emotion classification accuracy of the system reaches 95% on several test samples. The system is adapted to run on smart glasses, the emotion recognition results are immediately fed back to the screen of the smart glasses and displayed to the user in the form of an emotion label. The experimental results of system show that the system can achieve accurate emotion recognition, help users better understand the psychological state of the communication object, and improve the communication environment and quality. In addition, the system can also store communication records and emotion analysis results to help users understand their own communication performance and changes in others’ emotions, thus improving communication skills and abilities.

In the field of intelligent digital healthcare, Continuous-flow microfluidic biochip (CFMB) has become a research direction of widespread concern. CFMB integrates a large number of microvalves and large-scale microchannel networks into a single chip, enabling efficient execution of various biochemical protocols. However, as the scale of the chip increases, the routing task for CFMB becomes increasingly complex, and traditional manual routing is no longer sufficient to meet the requirements. Therefore, this paper proposes an automatic routing framework for CFMB based on Genetic algorithm (GA) and A* algorithms. Specifically, we adopt a two-stage A* algorithm to design the routing between modules, using the routing results obtained from the A* algorithm as the basis for evaluating the quality of solutions in the GA algorithm. Then, the GA algorithm is used to search for the optimal approximate solution in the solution space. Experimental results show that this method can reduce routing length and minimize routing crossings, thereby improving the parallel transmission speed of reagents on CFMB. This approach provides a feasible solution for large-scale automated routing of CFMB in the field of intelligent digital healthcare.

In situations where ordinary people and the hearing-impaired person need to communicate, it is possible that the average person does not know sign language and thus communication may be impaired, which means that a technology or device to assist communication is needed. Therefore, this study develops a new cloud sign language translation system on smart device based on the Browser/Server architecture, so that when the hearing-impaired person makes a sign language movement in front of the user who using the system with a smart device (e.g., smart glasses), the screen of the smart device will display the subtitle of the sign languages. We use MediaPipe to recognize and collect sign language action data from WLASL dataset and provide it to TensorFlow’s 1D-CNN deep learning model for training, so as to realize the sign language translation function. In the test phase, we invited five experimenters to test the sign language translation system ten times for each person, and the final average accuracy rate was 72%. Through such an interpreting system, it brings a convenient and efficient communication experience to the hearing-impaired person and people who need to communicate with the hearing-impaired person, and at the same time, it can also provide support for broader sign language research and application.

In recent years, numerous researchers have conducted in-depth studies and made significant progress in 2D Hand Pose Estimation (HPE) tasks on RGB images. However, the field of HPE in the context of infrared images has received limited attention. Due to the limited channel information and high correlation with temperature, models designed for RGB images may suffer from insufficient accuracy in infrared images. Our experiments reveal that the temperature distributions of the human hand in infrared images exhibit significant regularity. In this paper, we propose the Temperature-Based Hand Judgement Model (TB-HJM) that leverages this characteristic. During the training phase, a higher penalty is given when the predicted pose’s temperature distribution does not align with the actual temperature distribution, and vice versa. In the testing phase, TB-HJM is utilized to select a hand proposal that closely matches the temperature distribution as the final output. Additionally, to address the lack of visual information in infrared images, we use PBNHead and GCN Refine Module to merge structural information into the network to ensure model accuracy. Experimental results demonstrate that our model outperforms the benchmark model (HRNet) by 1.72% in terms of AUC and achieves an improvement of 0.6448 by reducing the EPE from 3.02 to 2.38, achieving state-of-the-art performance on our infrared hand dataset.

In smart communities, computer networks carry a large number of real-time computing tasks such as smart property, smart parking, smart home, etc., and these services are characterized by large data transmission and high concurrency, which require high real-time performance of the network. How to ensure the real-time network to reduce the task scheduling delay is the key problem to be solved when QoS optimization of smart community network. To this end, this paper deeply analyzes the characteristics of smart community task scheduling, firstly, establishes a community computing task scheduling model with computation time and computation cost as the optimization goal; then, proposes the optimization strategy of Chaotic Particle Swarm Algorithm for the stochastic nature of highly concurrent tasks, i.e., based on the basic algorithm of Particle Swarm to add the chaotic strategy in the initialization of the population and the optimization means of the adaptive factor in order to avoid falling into the local optimal and improve the optimization speed; finally, the time and cost overheads under different number of tasks are compared through simulation experiments, and the simulation results verify the effectiveness of the improved algorithm proposed in this paper in network QoS optimization.

Continuous-Flow Microfluidic Biochip (CFMB), with their integrated features, bring traditional biochemical experiments on a single chip to accomplish complex operations and reactions through precise control, efficient reactions and emerging ways of saving reagents. In the field of intelligent digital healthcare, CFMB have attracted a lot of attention. However, traditional manual design schemes can no longer meet the needs of increasingly complex chip architecture design. Therefore, this paper proposes an automated design method for resource binding and module placement of CFMB based on a list scheduling algorithm and an improved Simulated Annealing algorithm. Through the resource binding and scheduling design based on the list scheduling algorithm, an effective scheduling strategy is generated, which effectively improves the biochip execution efficiency. In addition, the improved Simulated Annealing algorithm solves the module placement problem in the biochip in a limited physical space. Compared with some benchmark algorithms, the experimental results demonstrate the effectiveness of the method in the biochip design process and provide a practical framework for further development of CFMB in the field of intelligent digital healthcare.

With the rapid development of broadband market services, Internet Service Providers (ISPs) strive to improve the end-to-end network quality of home broadband. However, faced with the opaque and unknown indoor network environment, traditional monitoring methods cannot meet the requirements, and new methods should be explored to identify problems on the user-side network to improve the perception and experience of users. In this method, FreeRadius and DaloRadius are deployed on the Linux server, user AAA attributes are collected by the Radius server, poor-quality routers and non-direct connection TV Set-Top Boxes (STB) are monitored with the help of Python's data processing capability, and the home networking environment of users is optimized to improve the perception and experience of users.

As a method of multi-agent system cooperation, multi-agent communication can help agents negotiate and adjust behavior decisions by exchanging information such as observation, intention, or experience during operation, improve the overall learning performance, and achieve their learning objectives. However, there are still some challenging problems in multi-agent communication. With the expansion of the multi-agent system scale, the global complete massive information will bring great resource overhead, and the introduction of redundant communication will lead to the difficulty of agent policy convergence, and affect the joint action and target completion. In addition, predefined communication structures have potential cooperation limitations in dynamic environments. In this paper, we introduce a dynamic communication model based on the graph convolution neural network called DCGN. Empirically, we show that DCGN can better cope with the dynamic update of tasks in the process of helping agents complete task information interaction, and can formulate more coordinated strategies than the existing methods.