Search Results (559)

The rapid growth of dockless electric scooter (e-scooter) sharing services has transformed short-distance urban mobility, offering convenience and sustainability benefits while amplifying challenges related to demand imbalance, fleet rebalancing, and spatial inequity. Accurate spatiotemporal demand prediction is therefore essential for optimizing resource allocation and supporting data-driven policy interventions. This study proposes a hybrid deep learning framework that integrates a Graph Convolutional Network (GCN) with a Gated Recurrent Unit (GRU) and community detection to enhance short-term prediction of e-scooter pick-up and drop-off demands. The Louvain algorithm is employed to partition urban areas into mobility-based communities, enabling the model to capture functional connectivity rather than relying solely on geographic proximity. Using real-world e-scooter trip data from Calgary, Canada, the model’s performance is evaluated against established baselines, including a Masked Fully Convolutional Network (MFCN) and conventional GRU architectures. Results show that the proposed approach achieves up to 11.8% improvement in mean absolute error (MAE) compared with the MFCN baseline and more robust generalization across temporal horizons. The findings demonstrate that integrating community structures into graph-based learning effectively captures complex urban dynamics, providing practical insights for sustainable micromobility operation and service deployment. Full article

In modern high-demand applications, such as real-time video communication, cloud gaming, and high-definition live streaming, achieving both superior transmission speed and high visual fidelity is paramount. However, unstable networks and packet loss remain major bottlenecks, making accurate and low-latency video error concealment a critical challenge. Traditional error control strategies, such as Forward Error Correction (FEC) and Automatic Repeat Request (ARQ), often introduce excessive latency or bandwidth overhead. Meanwhile, receiver-side concealment methods struggle under high motion or significant packet loss, motivating the exploration of predictive models. SimVP-v2, with its efficient convolutional architecture and Gated Spatiotemporal Attention (GSTA) mechanism, provides a strong baseline by reducing complexity and achieving competitive prediction performance. Despite its merits, SimVP-v2’s reliance on 2D convolutions for implicit temporal aggregation limits its capacity to capture complex motion trajectories and long-term dependencies. This often results in artifacts such as motion blur, detail loss, and accumulated errors. Furthermore, its single-modality design ignores the complementary contextual cues embedded in the audio stream. To overcome these issues, we propose A3DSimVP (Audio- and 3D-Enhanced SimVP-v2), which integrates explicit spatio-temporal modeling with multimodal feature fusion. Architecturally, we replace the 2D depthwise separable convolutions within the GSTA module with their 3D counterparts, introducing a redesigned GSTA-3D module that significantly improves motion coherence across frames. Additionally, an efficient audio–visual fusion strategy supplements visual features with contextual audio guidance, thereby enhancing the model’s robustness and perceptual realism. We validate the effectiveness of A3DSimVP’s improvements through extensive experiments on the KTH dataset. Our model achieves a PSNR of 27.35 dB, surpassing the 27.04 of the SimVP-v2 baseline. Concurrently, our improved A3DSimVP model reduces the loss metrics on the KTH dataset, achieving an MSE of 43.82 and an MAE of 385.73, both lower than the baseline. Crucially, our LPIPS metric is substantially lowered to 0.22. These data tangibly confirm that A3DSimVP significantly enhances both structural fidelity and perceptual quality while maintaining high predictive accuracy. Notably, A3DSimVP attains faster inference speeds than the baseline with only a marginal increase in computational overhead. These results establish A3DSimVP as an efficient and robust solution for latency-critical video applications. Full article

Unmanned aerial vehicles (UAVs) operating in dense urban environments often face link disruptions due to high mobility and interference. Reliable connectivity in such conditions requires advanced handover strategies. This paper presents a predictive-reactive Q-learning framework (PRQF) that optimizes handover decisions while sustaining throughput in dynamic heterogeneous urban networks. The framework combines an Extreme Gradient Boosting (XGBoost) classifier with a Q-learning agent through a probabilistic gating mechanism. UAVs follow a sinusoidal mobility model to ensure consistent and representative movement across experiments. Simulations using 3GPP-compliant Urban Macro (UMa) channel models in a 10 km × 10 km area show that PRQF achieves an average reduction of 84% in handovers at 100 km/h and 83% at 120 km/h, compared to the standard 3GPP A3 event-based handover method. PRQF also maintains a consistently high average throughput across all methods and speed scenarios. The results show better link stability and communication quality, demonstrating that the proposed framework is adaptable and scalable for reliable UAV communications in urban environments. Full article

As the demand for secure communication and data integrity in embedded and networked systems continues to grow, there is an increasing need for cryptographic solutions that provide robust security while efficiently using energy and hardware resources. Although software-based implementations of SHA-3 provide design flexibility, they often struggle to meet the performance and power limitations of constrained environments. This study introduces a hardware-accelerated SHA-3 solution tailored for the Xilinx Artix-7 FPGA. The architecture includes a fully pipelined Keccak-f [1600] core and incorporates design strategies such as selective loop unrolling, clock gating, and pipeline balancing to enhance overall efficiency. Developed in VHDL and synthesised using Vivado 2024.2.2, the design achieves a throughput of 1.35 Gbps at 210 MHz, with a power consumption of 0.94 W—yielding an energy efficiency of 1.44 Gbps/W. Validation using NIST SHA-3 vectors confirms its reliable performance, making it a promising candidate for secure embedded systems, including IoT platforms, edge devices, and real-time authentication applications. Full article

Directional frequency analysis and recording (DIFAR) is a widely used sonobuoy in modern underwater acoustic monitoring and surveillance. The sonobuoy is installed in the area of interest, collects underwater data, and transmits the data to nearby aircraft for data analysis. In this process, transmission of a large volume of raw data poses significant challenges due to limited communication bandwidth. To address this problem, existing studies on autoencoder-based methods have drastically reduced amounts of information to be transmitted with moderate data reconstruction errors. However, the information bottleneck inherent in these autoencoder-based methods often leads to significant fidelity degradation. To overcome these limitations, this paper proposes a novel autoencoder method focused on the reconstruction fidelity. The proposed method operates with two key components: Gated Fusion (GF), proven critical for effectively fusing multi-scale features, and Squeeze and Excitation (SE), an adaptive Channel Attention for feature refinement. Quantitative evaluations on a realistic simulated sonobuoy dataset demonstrate that the proposed model achieves up to a 90.36% reduction in spectral mean squared error for linear frequency modulation signals compared to the baseline. Full article

This paper presents fine-grained Field Programmable Gate Arrays (FPGA) architectures for the Advanced Encryption Standard (AES) MixColumns and InvMixColumns transformations, targeting improved performance and resource utilization. The proposed method reformulates these operations as boolean functions directly mapped onto FPGA Lookup-Table (LuT) primitives, replacing conventional xor-based arithmetic with memory-level computation. A custom MATLAB-R2019a-based pre-synthesis optimization algorithm performs algebraic simplification and shared subexpression extraction at the polynomial level of Galois Field $GF(2^8)$, reducing redundant logic memory. This architecture, LuT-level optimization minimizes the delay of the complex InvMixColumns stage and narrows the delay gap between encryption (1.305 ns) and decryption (1.854 ns), resulting in a more balanced and power-efficient AES pipeline. Hardware implementation on a Xilinx Virtex-5 FPGA confirms the efficiency of the design, demonstrating competitive performance compared to state-of-the-art FPGA realizations. Its fast performance and minimal hardware requirements make it well suited for real-time secure communication systems and embedded platforms with limited resources that need reliable bidirectional data processing. Full article

Financial time series are heterogeneous, nonstationary, and dispersed across institutions that cannot share raw data. While federated learning enables collaborative modeling under privacy constraints, fixed architectures struggle to accommodate cross-market drift and device-resource diversity; conversely, existing neural architecture search techniques presume centralized data and typically ignore communication, latency, and privacy budgets. This paper introduces FedRegNAS, a regime-aware federated NAS framework that jointly optimizes forecasting accuracy, communication cost, and on-device latency under user-level $(\epsilon ,\delta )$-differential privacy. FedRegNAS trains a shared temporal supernet composed of candidate operators (dilated temporal convolutions, gated recurrent units, and attention blocks) with regime-conditioned gating and lightweight market-aware personalization. Clients perform differentiable architecture updates locally via Gumbel-Softmax and mirror descent; the server aggregates architecture distributions through Dirichlet barycenters with participation-weighted trust, while model weights are combined by adaptive, staleness-robust federated averaging. A risk-sensitive objective emphasizes downside errors and integrates transaction-cost-aware profit terms. We further inject calibrated noise into architecture gradients to decouple privacy leakage from weight updates and schedule search-to-train phases to reduce communication. Across three real-world equity datasets, FedRegNAS improves directional accuracy by 3–7 percentage points and Sharpe ratio by 18–32%. Ablations highlight the importance of regime gating and barycentric aggregation, and analyses outline convergence of the architecture mirror-descent under standard smoothness assumptions. FedRegNAS yields adaptive, privacy-aware architectures that translate into materially better trading-relevant forecasts without centralizing data. Full article

Underwater marine and freshwater environments are vast and mysterious, but our ability to explore them is limited by the inflexibility and inconvenience of monitoring systems. To overcome this problem, in this work, we present a proof-of-concept deployment of a real-time Internet of Underwater Things (IoUT) using blue light-emitting-diode-based visible light communication (VLC). Pulse-amplitude modulation with four levels is employed. To relax the focus point and increase the received power, four avalanche photodiodes (APDs) are adopted. Moreover, to reduce the error rate, the convolutional code with constraint-7 is used, which is the simplest to implement. Encoding and decoding are implemented by a field-programmable gate array. The results are verified by experimental demonstration. A baud rate of 9600 is used, but, unfortunately, we only have a 2 m long tank. System performance is improved when the number of APDs is increased; we investigated the effects of up to four APDs. Notably, bit error-free data transmission can be achieved. Additionally, this method would make underwater monitoring very conventional and dependable, and low-cost real-time monitoring would be possible, with data shown on the Grafana dashboard tool. Full article

Differences in risk perception between frontline construction workers and managers can create communication barriers and lower the efficiency of safety management. In this study, we focused on frontline construction workers and managers and used event-related potentials (ERPs) to examine discrepancies in risk perception across two processes: hazard identification and risk judgment. During hazard identification, workers identified fewer hazards correctly than managers (p = 0.009 < 0.05). Managers also showed larger N200 amplitudes than workers (p = 0.040 < 0.05), which suggests that managers engaged conflict monitoring and inhibitory control more strongly. During risk judgment, workers responded more slowly than managers (p = 0.012 < 0.05). They also showed lower P100 (p = 0.026 < 0.05) and LPP amplitudes (p = 0.024 < 0.05), indicating weaker early visual–attentional gating and less sustained evaluative engagement with hazardous scenes. These patterns indicate that workers rely more on irrelevant information, whereas managers respond more sensitively to potential hazards. By revealing when and how role-based differences emerge, our findings offer a neurocognitive explanation for the persistent gap in risk perception and highlight specific targets for training. These insights can guide risk communication between managers and workers, extend research on risk-perception differences beyond self-report measures, and illustrate the value of ERP as a time-resolved tool for studying risk perception. Full article

Free-space optical (FSO) communication provides high-capacity and secure links but is strongly impaired by atmospheric turbulence, which induces multi-scale irradiance fluctuations. Traditional approaches such as adaptive optics, multi-aperture and multiple-input multiple-output FSO schemes offer limited robustness under rapidly varying turbulence, while statistical fading models such as log-normal and Gamma–Gamma cannot represent multi-scale temporal correlations. This work proposes a hybrid deep learning framework that explicitly separates high-frequency scintillation and low-frequency power drift through a conditional variational autoencoder and a bidirectional long short-term memory dual-branch architecture with an adaptive gating mechanism. Trained on OptiSystem-generated datasets, the model accurately reconstructs irradiance distributions and temporal dynamics. For model-assisted signal compensation, it achieves an average 79% bit-error-rate (BER) reduction across all simulated scenarios compared with conventional thresholding and Gamma–Gamma maximum a posteriori detection. Transfer learning further enables efficient adaptation to new turbulence conditions with minimal retraining. Experimental validation shows that the compensated BER approaches near-zero, yielding significant improvement over traditional detection. These results demonstrate an effective and adaptive solution for turbulence-impaired FSO links. Full article

In 2023, the bicycle industry in Taiwan reached a historic high. However, concerns about carbon emissions persist, particularly during the material acquisition and manufacturing stages of bicycle production. This study utilizes the Life Cycle Assessment (LCA) method, using SimaPro 9.5 for cradle-to-gate carbon emission data analysis. This study thoroughly examines the complete life cycle of a bicycle brake cable product through a carbon reduction evaluation tool, identifying carbon hotspots in the product’s life cycle. The data reveals that packaging accounts for the highest proportion of factory carbon emissions in the brake cable product analysis (34.42%), followed by the product’s casing (30.60%), with the leading materials being metal, plastic, and paper. Throughout the cradle-to-gate process, we collaborated with product developers to utilize the LCA carbon reduction evaluation tool to analyze the life cycle of the brake cable product. By aligning market and development needs, we supported manufacturers in identifying additional carbon reduction strategies at the material selection, mechanical design, and manufacturing process stages. These strategies include using natural raw materials, reducing packaging volume, developing lightweight products, and investing in integrated equipment. By implementing these measures, companies can reduce the product’s carbon footprint and enhance resource efficiency during production. This assessment tool serves as a communication bridge between designers and engineers, translating LCA quantitative data into references for design and management decision-making. It also functions as a simplified analytical tool for SMEs to conduct preliminary diagnosis of carbon emission hotspots and plan improvement directions, particularly suitable for manufacturers lacking consulting resources and carbon inventory capabilities. The research findings not only help companies integrate carbon reduction thinking early in product development, forming a closed-loop system of quantitative analysis and design actions, but also provide concrete references for Taiwan’s bicycle industry to promote supply chain collaboration, achieve green transformation, and meet global carbon reduction goals. Full article

This study addresses the technical gaps in current flood simulation for regulated lakes, such as insufficient accuracy in simulating complex gate and dam operation processes and low computational efficiency that fails to meet practical engineering needs. By employing an improved two-dimensional (2D) hydrodynamic model, it systematically analyzes flood control strategies for large regulated lakes. Using the August 2018 flood event for model validation, the final simulation results indicate that the current flood control capacity meets standards for 50-year floods (Nanyang 36.79 m, Weishan 35.99 m) but fails for 100-year floods, exceeding limits by 0.23 m (Nanyang 37.22 m) and 0.15 m (Weishan 36.64 m). The designed conditions reduce 100-year flood levels to 36.98 m and 36.47 m, respectively, achieving the required flood defense standard for 100-year events. The findings provide a quantitative framework for evaluating flood control capacity across different planning scenarios, which advances flood risk management and offers implementable insights for achieving sustainable water resource management in regulated lake basins globally. This, in turn, contributes directly to two United Nations Sustainable Development Goals (SDGs): enhancing human community safety and resilience (SDG 11: Sustainable Cities and Communities) through improved flood control engineering and operations, and strengthening climate adaptation (SDG 13: Climate Action) by boosting basin-wide resilience to extreme rainfall and flooding. Full article

This paper presents a Field Programmable Gate Array (FPGA)-implementable Frame Synchronizer that overcomes deficiencies of existing synchronizers in the space communications industry and provides a pipelined approach to achieve improved performance in latency, performance in the presence of noise, and streamlined implementation complexity. Unlike a soft decision synchronizer, magnitude (soft) bits are not required from the demodulation stage, and only the sign bit is used, reducing the complexity and signal counts between the transceiver and the synchronizer. Improved performance in noise can be achieved by introducing a small observation window surrounding the candidate Attached Sync Marker (ASM) window to uncorrelated data around the ASM. Further improvement in the presence of noise is achieved by using two ASMs, effectively doubling the ASM length of observation, but with no increase in the ASM pattern length and using existing predefined ASM patterns, thus remaining compliant with the Consultative Committee for Space Data Systems (CCSDS) standards. A parallel and pipelined implementation without a state machine eliminates latency from search, verify, lock, and flywheel states and reduces the effects of cycle slips of traditional flywheel state machine synchronizers. Full article

For humanoid robots to interact naturally with humans and seamlessly integrate into daily life, natural language serves as an essential communication medium. While recent advances in imitation learning have enabled robots to acquire complex motions through expert demonstration, traditional approaches often rely on rigid task specifications or single-modal inputs, limiting their ability to interpret high-level semantic instructions (e.g., natural language commands) or dynamically switch between actions. Directly translating natural language into executable control commands remains a significant challenge. To address this, we propose ToggleMimic, an end-to-end imitation learning framework that generates robotic motions from textual instructions, enabling language-driven multi-task control. In contrast to end-to-end methods that struggle with generalization or single-action models that lack flexibility, our ToggleMimic framework uniquely combines the following: (1) a two-stage policy distillation that efficiently bridges the sim-to-real gap, (2) a lightweight cross-attention mechanism for interpretable text-to-action mapping, and (3) a gating network that enhances robustness to linguistic variations. Extensive simulation and real-world experiments demonstrate the framework’s effectiveness, generalization capability, and robust text-guided control performance. This work establishes an efficient, interpretable, and scalable learning paradigm for cross-modal semantic-driven autonomous robot control. Full article

Reliable forecasts of dam releases are essential to anticipate downstream hydrological responses and to improve the operation of fluvial early warning systems. This study integrates an explicit release prediction module into a digital forecasting framework using the Lindoso–Touvedo hydropower cascade in northern Portugal as a case study. A data-driven approach couples short-term electricity price forecasts, obtained with a gated recurrent unit (GRU) neural network, with dam release forecasts generated by a Random Forest model and an LSTM model. The models (GRU and LSTM) were trained and validated on hourly data from November 2024 to April 2025 using a rolling 80/20 split. The GRU achieved R² = 0.93 and RMSE = 3.7 EUR/MWh for price prediction, while the resulting performance metrics confirm the high short-term skill of the LSTM model, with MAE = 4.23 m³ s⁻¹, RMSE = 9.96 m³ s⁻¹, and R² = 0.98. The surrogate Random Forest model reached R² = 0.91 and RMSE = 47 m³/s for 1 h discharge forecasts. Comparison tests confirmed the statistical advantage of the AI approach over empirical rules. Integrating the release forecasts into the Delft FEWS environment demonstrated the potential for real-time coupling between energy market information and hydrological forecasting. By improving forecast reliability and linking hydrological and energy domains, the framework supports safer communities, more efficient hydropower operation, and balanced river basin management, advancing the environmental, social, and economic pillars of sustainability and contributing to SDGs 7, 11, and 13. Full article