Traffic flow forecasting remains challenging because raw traffic flow observations often contain mixed temporal patterns, including slowly varying trends and fast local fluctuations. To address this issue, this paper proposes a Multivariate Empirical Mode Decomposition (MEMD)-guided dual-branch recurrent framework for multistep point forecasting. Specifically, MEMD is used as an alignment-preserving multivariate decomposition mechanism to obtain frequency-aligned components, which are then reconstructed into low-frequency trend and high-frequency residual components. The trend component is modeled by a Long Short-Term Memory (LSTM) branch to capture smooth long-term evolution, while the residual component is learned by a Bidirectional Gated Recurrent Unit (Bi-GRU) branch to characterize short-term oscillatory dynamics. A lightweight fusion head is then used to integrate the two branch-specific representations for final prediction. Experiments on PeMS04 and PeMS08, two traffic datasets derived from the California Department of Transportation Performance Measurement System, show that the proposed method achieves competitive performance across mean absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE), reaching 19.67/31.59/12.95% on PeMS04 and 15.51/24.43/9.86% on PeMS08. Compared with representative recent baselines, the proposed method achieves competitive results, with relative gains reaching 5.89% on PeMS04 and 5.35% on PeMS08 in selected metric-wise comparisons. These results indicate that MEMD-guided trend-residual representation learning can improve multistep traffic flow forecasting.

Aiming at the impact of obstacle blockage on task offloading performance in 6 G-era edge computing systems, this study proposes a partial computational offloading scheme assisted by a unmanned aerial vehicle (UAV)-mounted reconfigurable intelligent surface (RIS). First, under constraints of user transmission power, and task offloading ratio, a minimization problem for total user energy consumption is established. Second, this non-convex optimization problem is decomposed into three subproblems, and the proximal policy optimization (PPO) method from deep reinforcement learning is employed to determine the time slot allocation strategy. Finally, each training time step is treated as a solving instance, and an optimized solution is obtained based on alternating optimization methods. Simulation results demonstrate that the PPO-based algorithm trains rapidly and reduces the total user energy consumption.

In the realm of intelligent transportation and collaborative robotics, the accurate prediction of multi-agent trajectories has emerged as a critical challenge amid the ongoing advancements in autonomous driving and robotics technologies. Trajectory prediction aims to observe the historical motion trajectories of agents and predict their likely future locations. Despite the potential of deep generative models in this domain, popular approaches like Generative Adversarial Networks (GANs) and Variational AutoEncoders (VAEs) often grapple with concerns related to unsteady training, mode collapse, and subpar sample quality. In response to these challenges, we introduce DiffMATP, an innovative method for interaction-aware multi-agent trajectory prediction that leverages cutting-edge denoising diffusion models. DiffMATP integrates the notion of diffusion into trajectory prediction by modeling each agent's motion trajectory as a diffusion process. It employs reverse denoising to mitigate uncertainty during the prediction phase. Training DiffMATP involves variational inference on a parameterized Markov chain, optimizing the reweighted variational lower bound through mean square error minimization. Additionally, we incorporate temporal and spatial attention mechanisms into diffusion models to capture dynamic interactions among agents. Experiments conducted on pedestrian and highway vehicle datasets demonstrate that DiffMATP outperforms existing methods in terms of accuracy and generalization, which underscores the potential of denoising diffusion models in addressing the intricate challenges associated with multi-agent trajectory prediction.

Trajectory anomaly detection is becoming increasingly important in various fields such as smart cities, autonomous vehicles, and video surveillance. In recent years, pattern learning-based trajectory anomaly detection methods have emerged as a research hotspot, demonstrating strong modeling capabilities and flexibility. This review focuses on pattern learning-based trajectory anomaly detection technologies, providing a comprehensive and in-depth review of the research progress and emerging challenges in this field over the past decade. First, the unique characteristics of trajectory anomaly detection problems and the current research challenges are thoroughly analyzed. Then, existing pattern learning-based trajectory anomaly detection algorithms are classified. For each category of algorithm, its principles, representative literature, and an objective evaluation of its advantages and disadvantages are discussed. Furthermore, common anomaly detection evaluation techniques, open-source trajectory datasets, and the comparison of representative methods on related datasets are explored. Finally, the critical open issues in the field are summarized, and a prospective outlook on future research trends and potential solutions is provided, particularly regarding the potential and challenges of deep learning-based models in handling multifarious trajectory data.

To address the problem of service discontinuity caused by frequent user mobility in mobile edge computing, a microservice selection and deployment strategy was developed by integrating the cloud-edge speedup ratio and the edge gain index (EGI). Firstly, a microservice selection architecture was established based on containers and workflows, and a migration load optimization model incorporating service quality evaluation metrics was formulated. Then, a microservice deployment algorithm based on EGI was designed to achieve low-latency and high-hit-rate deployment. Furthermore, a microservice selection algorithm was proposed based on the cloud-edge speedup model and a dynamic cloud-edge migration strategy to handle cross-domain migration under user mobility. Simulation results demonstrate that the proposed app-roach improves service quality by approximately 23.2%, while reducing energy consumption and latency by 25.4% and 25.1%, respectively, compared to traditional methods.

Virtual traffic simulation is a crucial aspect within the autonomous driving field. To guarantee the safety of autonomous driving tests, it is imperative to generate traffic flows that closely approximate real-world scenarios. In this paper, we present an unsupervised approach grounded in traffic trajectory anomaly detection to quantitatively assess the authenticity of arbitrary trajectories. This method can also serve to compare the modeling capabilities of existing simulation algorithms. Initially, we utilize the original trajectory data as input and leverage the variational autoencoder (VAE) architecture to accomplish the spatiotemporal feature learning and reconstruction of the trajectory. Subsequently, trajectory anomaly detection is achieved by contrasting the disparity between the model’s output and the original data. Normal trajectories typically exhibit favorable continuity and smoothness, resulting in reconstruction outcomes that are closer to the original trajectories, whereas anomalous trajectories display the opposite characteristics. Finally, we introduce perceptual evaluation and derive the trajectory authenticity evaluation function via polynomial fitting. Experiments conducted on multiple trajectories demonstrate that our method can effectively measure the authenticity of both real-world and synthetic data. Notably, our method directly employs the original position information of the trajectory for model training, thereby enhancing the practicality and robustness of the approach.

A diversity-driven evolutionary strategy for microservice path generation in mobile edge computing is proposed to address load imbalance and user experience degradation caused by high request concurrency, resource heterogeneity, and centralized path selection. First, a backtracking search with diversity constraints is designed to overcome redundancy in conventional path generation, where a path-difference metric and pruning mechanism efficiently produce a structurally diverse, low-redundancy candidate path set. Then, a two-tier game-theoretic framework models user strategy evolution through evolutionary dynamics, guiding the shift from isolated decisions to coordinated distributions and mitigating resource contention across edge nodes. Finally, a discrete particle swarm optimization algorithm with recent-past position updating enables localized server pricing, steering requests toward underutilized nodes to relieve hotspot congestion. Extensive simulations demonstrate that the proposed strategy significantly improves path diversity, load balancing, and latency compared with state-of-the-art methods.

Considering that most pavement anomaly detection algorithms are difficult to play a stable role in data related to different distributions of pavement anomalies, this paper proposes a pavement anomaly detection algorithm based on multi-scale fusion time series information. The algorithm iteratively extracts autoregressive latent variables at different time series scales to quantify dynamic regularity information. Then iterative feature set is embedded into the pictographic subsequence learning module to rapidly extract most representative subsequence features through deep learning. The static statistical evaluation value is retained as one of the specific data to assist in identifying the sample, and the two angle data complement each other to maximize the feature information entropy. Using the SVM classifier with the global sequence alignment algorithm as the main kernel function, the data classification module detects anomalies, so that the algorithm can mine valuable feature attributes from multiple angles as much as possible while stably exerting its classification function. The experimental results show that the multi-scale fusion features capture more complete information about the temporal trajectory than the single latent variable feature information, suggesting that the proposed algorithm exhibits better comprehensive performance and generalization ability than unimproved machine learning algorithms and typical deep learning classifiers.

The emergence of multi-access edge computing (MEC) aims at extending cloud computing capabilities to the edge of the radio access network. As the large-scale internet of things (IoT) services are rapidly growing, a single edge infrastructure provider (EIP) may not be sufficient to handle the data traffic generated by these services. Most of the existing work addressed the computing resource shortage problem by optimizing tasks schedule, whereas others overcome such issue by placing computing resources on demand. However, when considering a multiple EIPs scenario, an urgent challenge is how to generate a coalition structure to maximize each EIP's gain with a suitable price for computing resource block corresponding to a container. To this end, we design a scheme of EIPs collaboration with a market price for containers under a scenario that considers a collection of service providers (SPs) with different budgets and several EIPs distributed in geographical locations. First, we bring in the net profit market price model to generate a more reasonable equilibrium price and select the optimal EIPs for each SP by a convex program. Then we use a mathematical model to maximize EIP's profits and form stable coalitions between EIPs by a distributed coalition formation algorithm. Numerical results demonstrate that our proposed collaborative scheme among EIPs enhances EIPs' gain and increases users' surplus.

To overcome the additional computation delay caused by straggler nodes in edge computing environments, a task encoding strategy based on systematic MDS code is proposed. The objective is to minimize the weighted sum of the total task processing delay and total energy consumption. A three-tier task allocation model consisting of network terminal devices, task encoding schedulers, and edge servers is established. Additionally, an optimal encoding parameter search algorithm is designed. The task encoding scheduler uses the encoding parameters determined by the algorithm to segment and encode the tasks, generating redundant subtasks. These subtasks are then allocated to edge servers for parallel processing, and the computation results are returned to the task encoding scheduler to recover the final results. Simulation results show that the proposed encoding scheme effectively reduces the total system cost compared to uncoded, replication-coded, and FRC schemes, while also demonstrating strong stability.

In the edge computing environment, there are risks such as communication channel risks and edge server failures, which can lead to a mismatch between the computing resources required for task processing and the resources allocated by the edge coalition. In response, a revenue forecasting method for the edge coalition and its member edge infrastructure provider (EIP) based on the game theory of uncertain coalition structures was proposed. Firstly, a resource scheduling model was constructed using a mixed integer linear programming method to maximize the revenue of the edge coalition. Secondly, a belief structure was introduced to characterize the probabilities of high, medium, low, and unknown scenarios for the coalition's revenue. Finally, the uncertain Owen value was used to estimate the interval revenue of the EIP in the coalition one time slot in advance. The simulation results show that the accuracy of this forecasting method under the two risks of channel risk and server failure is 91.25% and 82.5% respectively, with an average accuracy of 86.88%, achieving a relatively accurate forecast of the EIP’ revenue.

The development of high-throughput sequencing technology provides an opportunity to obtain multi-omics data for liver cancer,However,omics data often comes from different platforms and has different attributes, it has the characteristics of high feature dimension and small sample size. This will increase the overfitting of the model and the imbalance of categories,and the cross-platform integration analysis of omics data will challenge the traditional data analysis methods. In this regard, the Hierarchical Integrated Stacked Encoder (HI-SAE) is proposed.which can achieve deeper feature learning and data integration while reducing the differences caused by the characteristics of the data itself. Finally,the integrated feature expression is used to identify the subtype of liver cancer by softmax classifier. Experiments show that the classification accuracy when using Hi-SAE method for feature learning is 3.7% higher than that when using PCA, and 7.6% higher than that when using NMF.

Edge computing is considered a promising architecture for handling latency-sensitive and computationally intensive tasks. The lack of consideration for the timing of jobs and their unique topology in the existing research on task scheduling in mobile edge computing settings results in performance deterioration and underutilization of edge servers. In this paper, we provide a two-tier, lightweight offloading mechanism. by carefully taking into account: 1) the job topology, 2) the task urgency, and 3) the rivalry for server resources. Simulation findings show that, as compared to baseline techniques, our proposed approach considerably increases application completion rates while reducing average system completion delay by 66.7% and 37.6% under various user and edge server counts.

Aiming at the unreasonable cluster head selection in the classical LEACH algorithm, which causes uneven energy consumption and short network life cycle in wireless sensor networks, this paper proposes a Leach Protocol optimization based on Multi-hop with Distribution by Data Volume- Distance (LEACH-MP) algorithm. During network initialization, cluster head selection is performed with full consideration of node residual energy; the generation of very large clusters and very small clusters is reduced by homogenizing the number of cluster members in the cluster splitting phase; the data fused with cluster heads is passed to the base station in data transmission. Simulation experiments prove that the LEACH-MP algorithm can effectively extend the operation cycle of the sensor network and reduce the excessive consumption of node energy.

In this study, we investigate relay selection for latency-sensitive tasks in the context of NOMA (Non-Orthogonal Multiple Access) mode. Our objective is to enhance the stability of transmission and improve signal quality in satellite links, while also alleviating spectrum scarcity issues in ground base stations. To achieve this, we propose a three-layer relay architecture model in NOMA mode, utilizing distributed channel resource allocation through edge cloud servers to reduce decision complexity for satellites. Our aim is to minimize overall task transmission time, enhance task completion rates, and meet the minimum transmission rate requirements for ground base station users. By extracting user transmission information from ground base station configuration files and assigning priorities based on task characteristics, we employ iterative matching to determine relay selection for each satellite user. We conduct a detailed analysis of the impact of key factors on the performance of our proposed scheme and highlight the advantages of NOMA mode in relay systems.

Edge computing (EC) is a distributed computing paradigm that brings computation and data storage closer to the data sources. With the rapid development of EC, offloading scientific computing and data processing tasks from end devices to edge nodes (ENs) can satisfy these tasks’ requirements. However, a single EN with limited computing capacity is usually insufficient to handle these tasks. Hence, using a coalition structure (CS) of many ENs to handle multiple concurrent tasks in an EC environment becomes a feasible scheme. Still, many existing methods cannot be used directly in this scenario because of enormous CS solution space, low search speed, long-running time, poor optimal solution quality, etc. In response, we propose an arbitrary discrete political optimizer (ADPO) algorithm with a discrete recent-past position updating strategy to explore the potential CS space. Unlike other heuristic algorithms, ADPO further improves the better solutions by interacting with each other in the parliamentary affairs phase. After that, we formulate an integer programming (IP) optimization model to alleviate the low search speed or long time consuming. Finally, extensive experiments demonstrate that ADPO is superior to the existing heuristic algorithm. In addition, the IP model excels in exiting algorithms at the running time (milliseconds level) and resource occupied ratio.

A discrete recent past-position updating strategy based m-ary discrete particle swarm optimization (MDPSO-DRPPUS) algorithm was proposed for the problem of large search space and low efficiency when solving the optimal coalition structure.First, the coalition structure with index-based was coded.Then, the multi-objective optimization problem was transformed into an eigenvalue function of the coalition structure.Finally, the optimal coalition structure was searched by using the MDPSO-DRPPUS algorithm.Experiments show that compared with the m-ary discrete particle swarm optimization (MDPSO) algorithm and genetic algorithm (GA), the proposed algorithm dramatically reduces the average running time, and improves the efficiency and equilibrium of the coalition structure and task completion efficiency of edge nodes.

Edge computing empowers the IoV to achieve performance requirements such as low latency and high computational load for in-vehicle services. However, the driving of vehicles is random and unevenly distributed, causing problems such as unbalanced load of edge servers and low edge resource utilization. Therefore, in this article, based on the vehicle trajectories, the edge resource allocation algorithm and load balancing algorithm are used to obtain the load prediction value of the edge server and then calculate the optimal edge resource quantity in order to reduce the resource idleness as much as possible. The experiments demonstrate that the application of the edge resource allocation algorithm and load balancing algorithm based on vehicle trajectory significantly reduces the blocking rate of edge resource requests by vehicles and improves the benefits of the overall IoV edge system.

Based on three tier heterogeneous network, we examine the problem of energy consumption optimization with latency constraints, establishing a hierarchical computing model with multiple layers under heterogeneous network, each layer representing its individual computing model. An offloading strategy will be optimized to reduce energy consumption while ensuring latency constraints. Lyapunov optimization theory will be utilized along with a heuristic algorithm called Coral Reef Optimization (CRO) to analyze the current workload and available computing capacity of edge and cloud servers, making an efficient compute offloading choice reduces the energy used by cloud servers while still adhering to latency restrictions.. Simulation result shows that compared to cloud-only and edge-ward strategies, the proposed strategy shows improvements in term of energy reduction as well as execution time.

Deep learning technology is widely used in computer vision.Generally,a large amount of data is used to train the model weights in deep learning,so as to obtain a model with higher accuracy.However,massive data and complex model structures require more calculating resources.Since people generally can only carry and use mobile and portable devices in application scenarios,neural networks have limitations in terms of calculating resources,size and power consumption. Therefore,the efficient lightweight model MobileNet is used as the basic network in this study for optimization.First,the accuracy of the MobileNet model is improved by adding methods such as the convolutional block attention module (CBAM)and expansion convolution.Then,the MobileNet model is compressed by using pruning and weight quantization algorithms based on weight size.Afterwards,methods such as Python crawlers and data augmentation are employed to create a garbage classification data set.Based on the above model optimization strategy,the garbage classification mobile terminal application is deployed on mobile phones and raspberry pies,realizing completing the garbage classification task more conveniently.