In large-scale urban areas, edge computing, with flexible and low-latency services, enriches various pioneering applications such as the internet of vehicles (IoVs) and smart cities. These location-sensitive applications raise a critical concern, i.e., the mismatch between the spatial distributions of computing requirements and computing capacities. And this mismatch gives a huge challenge for load balancing among edge servers. However, existing approaches do not account for this spatial unevenness, which undermines the high-quality implementations of urban edge computing systems. Regarding this load balancing problem, we propose a novel Power Diagram based Edge Balancing (PDEB) approach, pursuing computing capacities self-adapted with computing requirements via the power diagram. This paper makes three key contributions. First, we propose a distribution modeling framework that formulates spatial mismatch as an optimization problem. Second, we introduce a decentralized power diagram constructed from modeled distributions, serving as the mathematical foundation for decentralized edge coordination. Third, we develop a structured pairing and scheduling strategy based on the power diagram to proactively redistribute load across edge servers. Experimentally, PDEB achieves$42.78\%$better load balance and$16.69\%$lower queuing delay than leading baselines, validating its theoretical and practical advantages.

Graphical User Interfaces (GUIs) play a crucial role in facilitating user-computer interactions, making them an essential focus of research. However, current automated GUI agents face significant challenges in effectively associating task implementations with specific visual elements, and the resolution constraints of Vision-Language Models (VLMs) also limit the richness of visual information. To this end, we propose a novel multi-modal agent named MT-Agent, which enhances both textual and visual input modalities to enable the model to perceive visual elements in GUIs more effectively. Specifically, Textual Modality Enhancement improves the semantic richness of input text by capturing task-specific details via an external VLM, while Visual Modality Enhancement incorporates fine-grained visual details to better represent critical GUI elements. In addition, we introduce an innovative text-guided directional feature fusion mechanism, which leverages enriched text features to guide the integration with visual information. In experiments, MT-Agent demonstrated exceptional performance on AITZ dataset, achieving an action type prediction accuracy of 84.80% and a step prediction accuracy of 58.07%, surpassing previous state-of-the-art models. Furthermore, on the GUI Odyssey benchmark, MT-Agent achieves performance comparable to previous state-of-the-art models while using only about 1/20 of their trainable parameters. Our codes, demos, and relevant data will be released to facilitate further research and validation within the scientific community.

With the rapid development of Mobile Edge Computing (MEC), various heterogeneous applications have being deployed on edge servers in close proximity to end-users for the low-latency responses. In this circumstance, since the resources on edge servers are limited, it is critical to deploy these applications on suitable edge servers. However, due to the heterogeneity of the applications and the mobility of end-users in real MEC circumstances, the requests each edge server received may undergo temporal fluctuations in both views of quantity and type. In other words, it is crucial to re-deploy these heterogeneous applications to match these dynamic circumstances, instead of permanent deployments without adjustments. Nevertheless, frequent re-deployment causes service interruptions and resource wastage, leading to system instability. Existing approaches struggle to handle redeployment effectively in heterogeneous, dynamic, and stability-critical MEC environments. In this paper, we first formulate the Edge Application Re-Deployment problem on the basis of constrained multi-objective optimization and prove its$\mathcal {NP}$-hardness. Then we propose an optimal re-deployment approach based on the Integer Programming technique for small-scale edge application re-deployment scenarios. And we also propose a Decompose-Solve-Merge approximation approach which balances the effectiveness and efficiency with a configurable parameter for large-scale scenarios. Extensive experiments on a real-world data set evaluate our novel approaches against four existing representative approaches. Additionally, we perform the ablation experiment to validate the effectiveness of our approaches and explore the impact of configurable parameter on the performance. The results show the superior performance of our approaches on re-deployment in terms of heterogeneous, dynamic, and stability.

As the capabilities of smart sensing and mobile technologies continue to evolve and expand, storing diverse sensor data on smartphones and cloud servers becomes increasingly challenging. Effective data compression is crucial to alleviate these storage pressures. Compressed sensing (CS) offers a promising approach, but traditional CS methods often struggle with the unique characteristics of sensor data-like variability, dynamic changes, and different sampling rates-leading to slow processing and poor reconstruction quality. To address these issues, we developed Mob-ISTA-1DNet, an innovative CS framework that integrates deep learning with the iterative shrinkage-thresholding algorithm (ISTA) to adaptively compress and reconstruct smartphone sensor data. This framework is designed to manage the complexities of smartphone sensor data, ensuring high-quality reconstruction across diverse conditions. We developed a mobile application to collect data from 30 volunteers over one month, including accelerometer, gyroscope, barometer, and other sensor measurements. Comparative analysis reveals that Mob-ISTA-1DNet not only enhances reconstruction accuracy but also significantly reduces processing time, consistently outperforming other methods in various scenarios.

Although micro-mobility has become a popular and indispensable mode of transportation in recent years, it has also introduced a large number of traffic accidents. Timely tracking and predicting the maneuvers hold the potential to prevent accidents through prompt warnings and interventions. However, the open and simple structure of micro-mobility makes it hard to install sophisticated infrastructures for maneuver prediction. In this paper, we argue that the micro-mobility body dynamics provide sufficient information for maneuver prediction. Our preliminary study suggests that micro-mobility body dynamic patterns appear beforehand and exhibit the correlation with steering maneuvers. We accordingly present RideGuard, which leverages a built-in Inertial Measurement Unit on smartphones to achieve the prediction of steering maneuvers. Through a dual-stream CNN deep learning architecture, RideGuard effectively captures complex patterns and feature relationships from the time and frequency domain. Our extensive real-traffic experiments involving 20 participants demonstrate the superiority of RideGuard: employing a 3s detection window, RideGuard attains a minimum of 94% precision in maneuver prediction with a 5s prediction time gap. The low-cost and rapid response feature of RideGuard enables feasible deployment and promotes safer riding practices. Additionally, we open-source our well-labeled dataset to facilitate further research.

Edge computing deploys edges close to end-users to provide highly accessible resources and latency-sensitive services. It is invaluable for urban crowd/hazard management services, e.g., real-time dynamic route planning and hazard monitoring/analysis, etc. However, in such scenarios, various types of urban hazards jeopardize the usability of edge servers. Worsely, these hazards could be integrated, like gas fires caused by urban earthquakes. In this regard, the formulation of usability risks that servers face is intractable due to the complexity, incomplete real-time data and insufficient expert knowledge of these integrated hazards. Therefore, we innovatively define the usability risks asServer Hazard Riskmodel from the view of the spatial data field by utilizingInformation Diffusion techniquewhich can overcome the adverse conditions above. Then we involve it to formulate theServer Hazard Risk User Allocation(SR-UA) problem, and analyze three typical solutions from the perspective of optimality and efficiency, which are the Lexicographic Goal Programming approach (SR-UA-LGP), the Approximation approach (SR-UA-A) and the Particle Swarm Optimization-based approach (SR-UA-PSO). The extensive experiments based on two real-world datasets illustrate the superior performance of our model and solutions.

Although micro-mobility brings convenience to modern cities, they also cause various social problems, such as traffic accidents, casualties, and substantial economic losses. Wearing protective equipment has become the primary recommendation for safe riding. However, passive protection cannot prevent the occurrence of accidents. Thus, timely predicting the rider's maneuver is essential for active protection and providing more time to avoid potential accidents from happening. Through the qualitative study, we argue that we can use the rider's head dynamic as an information source to predict the rider's following maneuvers. We accordingly present HeadMon, a riding maneuver prediction system for safe riding. HeadMon utilizes the head dynamics of a rider by installing an inertial measurement unit on the helmet. It uses the extracted head dynamics features as the input of the deep learning architecture to achieve prediction. We implemented the HeadMon prototype on Android smartphone as a proof of concept. Through comprehensive experiments with 20 participants, the result demonstrates the excellent performance of HeadMon: not only could it achieve an overall precision of at least 85% for maneuver prediction under a 4s prediction time gap, but it also could keep a high accuracy under a low sampling rate. The low-cost feature of HeadMon allows it to be readily deployable and towards more safety riding.

As intelligent sensing and smartphone technologies have progressed, a huge amount of highly heterogeneous data have come to be stored in smartphones and uploaded to servers for analysis on a daily basis. This has led to vast storage overheads for users and companies. Hence, data compression becomes the most efficient strategy for suppressing the increase in storage overhead. Compressed sensing (CS) technology is one approach to compressing data, but traditional CS-based algorithms are significantly time-consuming and have low reconstruction performance. In light of these drawbacks, this paper proposes a compressed sensing framework that instead takes advantage of the low time cost and adaptive learning capability of deep learning methods, wherein a convolutional neural network (CNN) is used for compressing and reconstructing acceleration data. Our experiments with actual smartphone acceleration data show that the proposed method dramatically improves the reconstruction performance with very little reconstruction time compared with traditional compressed sensing methods.

This paper introduces mind evolutionary algorithm (MEA) into the application of short-term traffic flow prediction, and proposes a short-term traffic flow prediction model of wavelet neural network based on mind evolutionary algorithm (MEA-WNN). The optimal connection weight and wavelet parameters of wavelet neural network (WNN) are searched globally by MEA, and the convergence capacity of wavelet neural network is improved. The experimental data show that, compared with the prediction model of the traditional WNN and the WNN based on genetic algorithm (GA-WNN), the prediction model of MEA-WNN has higher global prediction accuracy.

This paper presents a method to integrate Wi-Fi networks with global positioning system (GPS) and dead reckoning (DR) system for tracking a vehicle in the urban areas by using the federated form of unscented Kalman filer (UKF) and Kalman filer (KF). Due to non-line-of-sight conditions and multipath propagation environments, the estimation of the GPS/DR integrated positioning system lacks acceptable accuracy for demanding applications, such as continuous vehicle tracking. Comparing with GPS, a fingerprinting positioning method based on Wi-Fi signal strength observations could continuously log location information without visibility to the sky. Numerical results show the improvement of the algorithm proposed in the paper.