Hong Kong University of Science and Technology
Hong Kong University of Science and Technology Institutional RepositoryNot a member yet
162821 research outputs found
Sort by
Six Switch Five-Level Boost-Type ANPC Inverter with Full DC Source Utilization
No full textThis article proposes a five-level (5L) boost active neutral point clamped (BANPC) switched capacitor (SC) inverter topology with reduced device count. Unlike conventional ANPC inverters, the proposed topology utilizes the full DC-link voltage, eliminating the need for additional boost converter stages. The SC impulse charging current is suppressed by approximately 20% even without employing a soft charging technique. To further minimize the inrush current, a soft charging technique is employed. As a result, the charging current is limited to two times the load current. Moreover, the proposed DC-link and switched capacitor voltages are self-balanced and do not require additional balancing circuits. Design considerations and a comparative analysis with other recent ANPC topologies are discussed, demonstrating that the proposed BANPC is superior in achieving dual operation with the least device count. Experimental results for a 500 W system are presented to validate the effectiveness and feasibility of the proposed BANPC inverter.</p
Discriminatory order assignment and payment-setting of on-demand food-delivery platforms: A multi-action and multi-agent reinforcement learning framework
No full textThis paper studies the discriminatory order assignment and payment-setting strategies for on-demand food-delivery platforms. We consider an on-demand food-delivery platform that coordinates customers, couriers, and restaurants to maximize the profit. It determines how to bundle orders, assign orders to couriers, and set payments to couriers in real-time. These decisions are made in a personalized manner, depending on the historical data collected from each of the couriers, such as the order acceptance and rejection rates under distinct scenarios of order assignment and payment values. A Markov Decision Process is formulated for the courier, capturing the decisions of the platform (including differentiated order assignment/bundling strategies and the discriminatory payment-settings decisions) while considering its dependence on the personalized work-related data of each individual courier. To derive the optimal policies, we propose a novel multi-action and multi-agent deep reinforcement learning framework, where a double Deep Q-Network is employed to develop discrete order assignment strategies, and double Proximal Policy Optimization is utilized to determine continuous payment decisions. Within this learning framework, we introduce a novel neural network architecture that leverages the Query-Key attention mechanism to transform multiplicative time complexities into additive computation complexity for order assignment, and we adopt a variable-length Bi-LSTM module that compresses variable-length order sequence into a fixed-dimensional feature space to enhance scalability. The proposed neural network and algorithmic framework was validated in a case study using real-world food-delivery data from Hong Kong. By comparing the proposed method with a vanilla MLP-based neural network architecture, we find that the proposed neural network architecture significantly enhances platform performance: it increases the number of orders served by 5.25%, reduces platform expenses by 10%, and improves the overall reward of the platform by over 50%. Additionally, our results reveal that couriers with higher order rejection rates receive more orders during peak hours but earn lower wages. This counterintuitive finding is attributed to a strategic approach by the platform to differentiate order allocation: instead of simply allocating fewer orders to couriers with higher rejection rates, the platform preferentially assigns longer-distance trips to couriers with a higher likelihood of order acceptance. These findings expose the implicit biases in the discriminatory algorithms used by the profit-maximizing platform and highlight potential areas for governmental regulatory intervention. The code of this paper is provided at https://github.com/RS2002/Discriminatory-Food-Delivery .</p
The "last-mile efforts" of subseasonal prediction and services for climate resilience and sustainability: review and outlook
No full textSubseasonal predictions from 2 weeks to 2 months have made significant advancements over the past decade, driven by progress in physical understanding, climate modeling, computational capabilities, and artificial intelligence (AI). These predictions are increasingly in demand due to their potential to provide stakeholders with adequate lead time for effective disaster adaptation, mitigation, and resource management. However, there remain critical gaps in the engagement between prediction providers and service users. Providers often lack insight into the specific needs of users and do not have transferrable strategies to build trust through tailored evaluations and clear confidence levels, which often results in repeatedly devising approaches for each provider–user interaction. Further, users frequently struggle to interpret predictions and are hesitant to make decisions based on these uncertain outcomes. This paper attempts to make “last-mile efforts” by reviewing relevant literature, operational systems, and the most informative communications and engagement strategies with key sectors. It proposes a preliminary framework to standardize the approach for provider–user interaction in the context of subseasonal prediction and services, with potential applicability and extension to seamless prediction systems in the future. Lastly, we underscore future directions for subseasonal predictions, emphasizing the integration of dynamic climate modeling and AI-driven enhancements with large ensemble techniques to improve both reliability and confidence. This review is part of the United Nations Educational Scientific and Cultural Organization (UNESCO) International Decade of Sciences for Sustainable Development (2024–33) and contributes to the Seamless Prediction and Services for Sustainable Natural and Built Environment (SEPRESS) Program (2025–32), an initiative endorsed under this global framework.</p
An online forecasting-based fine-tuning pipeline for time-series anomaly prediction
No full textTime-series anomaly detection is critical for numerous real-world applications and has been extensively studied. However, existing methods are typically designed to identify anomalies within a complete time series. In other words, they rely on access to ground truth data to calculate anomaly scores and distinguish anomalous data from normal patterns. This reliance limits their applicability in scenarios where predicting future anomalies is required, as the ground truth is inherently unavailable. To address this gap, we introduce the concept of Time-Series Anomaly Prediction (TSAP), which focuses on forecasting the occurrence and progression of anomalies in time series simultaneously without relying on ground truth. In this paper, we propose a novel exemplar-based pre-training and fine-tuning pipeline tailored to this task, based on recent achievements in online time-series forecasting techniques. The pipeline begins with an offline pre-training phase, where a deep learning model is trained to capture the underlying temporal correlations in time-series data. During the online fine-tuning stage, a three-step process is employed to predict the timing and evolution of anomalies. This process includes prediction and anomaly detection, motif search for similar patterns, and fine-tuning using exemplars. These steps are repeated as new data arrives. We evaluate the proposed method against state-of-the-art approaches from various relevant categories on both real-world and synthetic datasets. Experimental results show that the proposed method improves anomaly detection accuracy by up to 53.8% in terms of F1 score and enhances time-series forecasting accuracy during and after anomaly periods by up to 82.4% and 49.1% in terms of MSE. Through analysing the results, we prove the proposed method's effectiveness in addressing the new TSAP tasks, which are incapable of being handled by current time-series anomaly detection or online time-series forecasting methods.</p
Towards Green Manufacturing of Thermoplastic Composites: Quantifying Material-Process-Structure-Property Relationships and Data-Driven Design
No full textAngular-momentum enhanced non-hourglass formulation for SPH solid dynamics
No full textUpdated Lagrangian smoothed particle hydrodynamics (SPH) for solid dynamics is often plagued by numerical instabilities, particularly hourglass modes that produce unphysical zigzag patterns. While recent essentially non-hourglass (SPH-ENOG) and generalized non-hourglass (SPH-GNOG) formulations have improved stability, they suffer from poor angular momentum conservation, limiting their accuracy in rotational problems. To overcome this, this paper presents two angular-momentum enhanced non-hourglass formulations. First, we enhance the SPH-ENOG method with rotation matrices derived from Rodrigues’ formula, creating SPH-ENOG-A for elastic materials, which explicitly accounts for rigid rotations during time integration, thereby significantly enhancing angular momentum conservation. To strictly enforce linear momentum conservation, the average of the rotation matrices is computed and applied to each particle. We then extend this approach to reformulate the corrective term in SPH-GNOG, yielding SPH-GNOG-A—a unified method for both elastic and plastic materials that not only improves angular momentum conservation but also eliminates prior dependencies on material-specific coefficients. Validated against elastic (oscillating plates, spinning solids) and plastic (Taylor bars, high-velocity impacts) benchmarks, our methods retain the hourglass-free stability, convergence, and accuracy of their predecessors while achieving a significant leap in angular momentum conservation.</p
Unravelling the evolution of nickel-catalyzed C–O bond activation with data-driven strategies
No full textSince the 1970s, nickel has proven to be an exceptionally efficient catalyst for cross-coupling reactions, particularly in the activation of C–O bonds, which serves as an environmentally friendly alternative to organic halides. The relentless exploration by chemists of the synthetic methodologies and mechanisms of this field has progressively fostered the emergence of an increasingly mature yet intricate discipline. Despite its apparent complexity, the core patterns remain hidden within some significant works. The development of large language models (LLMs) has provided unprecedented opportunities to navigate this complex landscape and uncover hidden patterns. Here, we introduce GPT-NiCOBot, a modular platform that integrates LLMs with chemistry-specific tools to autonomously extract reactions and identify key patterns in reagents and catalysts from peer-reviewed papers. Moreover, by combining the core citation network with in-depth chemical knowledge, this platform constructs a more effective and comprehensive research assistance framework. This system demonstrates the potential of LLMs to accelerate research in nickel catalysis and suggests broader applications in other chemical subfields.</p
Gold-Catalyzed Carbonyl Release and its Adaptation for Prodrug Therapy Using Multivalent Lectin-Directed Artificial Metalloenzymes
No full textIn the framework of developing artificial metalloenzyme (ArM) prodrug therapies, two main factors need to be considered; the cancer targeting capabilities of the ArM biocatalyst and the bioorthogonal prodrug activation mechanism. In this study, both these aspects were investigated to develop an example of an anticancer ArM prodrug strategy. To address targeting, the concept of multivalent lectin-directed artificial metalloenzymes was established using a Halotag-PduU-ACG lectin fusion protein (HtPA) functionalized with a gold catalyst. Acting through multivalent binding of hexameric lectin complexes (caused by PduU oligomerization), selective binding to sialic acid-rich cancer cells was proven. To address prodrug activation, the propargylbenzoxime (PBO) group was developed to undergo gold-catalyzed hydroamination, followed by spontaneous N–O bond cleavage to release carbonyl functional groups under mild and physiological conditions. Further adaptation of the PBO group was also explored so that carbonyl release could elicit the synthesis of indole-containing molecules. HtPA-based artificial metalloenzymes were then subsequently applied in cell assays for the activation of a PBO-based prodrug to highlight this alternative approach of an ArM prodrug therapy.</p
Green synthesis of mechanically robust and antibacterial chitosan colloidal hydrogels under near-neutral conditions for wound dressing
No full textChitosan hydrogels are promising biopolymer systems for wound dressing applications, but their effectiveness is often limited by insufficient mechanical properties. Although chemical crosslinking can improve these properties, it typically involves the use of toxic compounds. Additionally, the solubility of chitosan in acidic conditions restricts its broader applications. Herein, we present a facile method using pH adjustment to near-neutral levels (6.4–6.6) followed by shear homogenization to self-assemble chitosan colloidal particles (CCPs) into a robust hydrogel. Silver nanoparticles (AgNPs) are generated in situ via a novel diffusion-controlled, chitosan-mediated reduction of silver nitrate (AgNO3). The resulting CCP-AgNP hydrogel exhibits an anisotropic layered network and excellent mechanical properties, with AgNPs uniformly dispersed throughout the matrix. The compressive fracture stress and strain reached 0.7–1.35 MPa and 50 %–75 %, respectively, likely due to the coordination between the chitosan and silver species. The CCP-AgNP hydrogel exhibited strong antibacterial activity against both Staphylococcus aureus (S. aureus) and Escherichia coli (E. coli). In an in vivo S. aureus-infected full-thickness wound healing model, the hydrogel significantly enhanced wound healing with no cytotoxicity detected as shown by cell viability assays. These findings suggest that CCP-AgNP hydrogels, prepared through this simple, environmentally friendly method, have significant potential as robust and effective wound dressings.</p
Thermal induced anisotropic deformation of DCPD gels during frontal polymerization
No full textAs an emerging low-energy-consumption and rapid-processing technique, frontal polymerization (FP) has garnered increasing attention due to its significant potential in free-standing three-dimensional printing, advanced composite manufacturing etc. However, effects from thermal expansion, chemical shrinkage, and nonuniform reaction propagation during the FP process inevitably generate gradient deformations at the FP front, adversely affecting the mechanical properties of the final processed components. To elucidate the intrinsic deformation mechanisms of polymer materials during FP and optimize their mechanical performance after polymerization, we employ a representative dicyclopentadiene (DCPD) gel system for investigation. In this study, simultaneous temperature and strain measurement revealed anisotropic deformation patterns of DCPD gels during FP. Meanwhile, experimental and numerical studies investigated the strain evolution of DCPD gels during both single and multi-point initiation of FP, revealing that the maximum compressive strain in uncured regions gradually decreased with increasing pre-curing degree but increased with rising width. Finally, mechanical characterization results show that, as the pre-curing degree increases, both compressive and tensile performances of polydicyclopentadiene are progressively enhanced, with yield strengths under tension and compression at 77 K significantly exceeding those at room temperature. These findings provide an experimental and theoretical foundation for high-performance material fabrication based on FP of gel-state systems.</p