Search Results (565)

Geometric problem-solving has always been a great challenge in the field of deductive reasoning and artificial intelligence. Symmetry is a defining characteristic of geometric shapes and properties. Consequently, the application of symmetry principles to geometric reasoning arises as a natural choice. To address the efficiency degradation and limited generalization, we propose FGeo-ISRL, a neural-symbolic inverse search framework whose core is the synergistic integration of a task-fine-tuned large language model and Monte Carlo Tree Search. Under the formal framework of FormalGeo, geometric theorems are iteratively applied starting from the given conditions and the target conclusion, in order to infer the necessary supporting premises. The large language model simultaneously serves as a policy network and a value network, guiding theorem application decisions and evaluating intermediate proof states, whereas the Monte Carlo Tree Search performs structured exploration over the state space, both training for policy refinement and inference for online search. The reinforcement learning agent is trained with a hybrid reward scheme, combining immediate feedback from the value difference and a sparse success reward. Experiments demonstrate the effectiveness and correctness of FGeo-ISRL. It not only achieves a Single-Step Theorem Accuracy of 90.2% and a Geometric Problem-Solving Accuracy of 83.14%, but also ensures that every step of the proof process remains readable, verifiable, and traceable. Full article

Dynamic fee market mechanisms, exemplified by EIP-1559, have been extensively studied for Ethereum’s 12 s block environment but remain uncharacterized at sub-second timescales. We present an agent-based simulation study of an EIP-1559 adaptation for Injective, a Layer 1 blockchain combining native EVM compatibility with CometBFT consensus, operating at 600 ms block times. Across twelve simulation runs (four parameter configurations × three demand scenarios), our analysis yields three findings: (1) temporal smoothing mechanisms (MA-25, 25-block trailing average) produce mixed effects in sub-second environments with up to 47% basefee overshoot during spam attacks and slight smoothing elsewhere, making per-block mechanisms preferable for consistent performance; (2) transitioning from 150M (66.66% target) to 300M (50% target) configuration reduces peak fees by 31% during variable demand; during spam attacks, the 300M configuration peaks 32% higher but recovers faster with block capacity as the primary driver for spam throughput; and (3) per-block mechanisms establish initial spam barriers within 17–32 s versus Ethereum’s 4–6 min, economically justifying lower minimum fees. We provide the first systematic sub-second EIP-1559 analysis and a parameter optimization framework for high-throughput chains. With proper tuning, dynamic fee mechanisms are compatible with high-throughput architectures. Full article

Human-centered entrepreneurship considers employee well-being and uses the Sustainable Development Goals as its fundamental pillars. However, existing research predominantly focuses on institutional interventions and fails to provide integrated intelligent solutions for tackling human–machine collaboration issues in the context of digital transformation. Large language models (LLMs) offer potential for affective computing and personalized support, but face critical gaps in ethical governance, privacy protection, and real-time risk intervention in sensitive entrepreneurial contexts. Our proposed Human-Centered Entrepreneurial Intelligent Agent (HCEA) framework achieves the unified optimization of task utility, empathetic expression, and ethical security by integrating a large language model core fine-tuned via a multi-objective hybrid loss function and a cluster of task-specialized intelligent agents. HCEA integrates retrieval-enhanced generation to ensure suggestion accuracy, a hierarchical data governance system for sensitivity-based privacy protection, and an independent risk detection module for real-time intervention and referral. We build the framework by constructing a hybrid entrepreneurial dataset, design the multi-agent architecture of decision support, emotion understanding and ethical risk tracking, and empirically evaluate both comparisons and ablation experiments. The results demonstrate that HCEA outperforms five baseline models across six key metrics, including entrepreneurship guidance relevance, emotion recognition, and high-risk recall. This study contributes to the intersection of digital transformation and sustainable entrepreneurship by providing a technically feasible, ethically grounded intelligent framework that empowers enterprises to reconcile efficiency with human-centric values, advancing SDG 8 (decent work and economic growth) and SDG 9 (industry, innovation, and infrastructure). Full article

In recent years, DC-DC partial power converters (PPCs) have become increasingly important in fast-charging architectures for electric vehicles (EVs). Their key feature is that only a fraction of the energy delivered to the battery is processed by the PPC, while the rest is transferred directly, bypassing the conversion stage. This reduces DC-DC conversion losses and improves overall charging efficiency. However, the nonlinear dynamics of these converters can limit performance, especially with model-based controllers such as proportional–integral (PI) controllers. This paper proposes a data-driven control framework for EV fast-charging stations using a DC-DC PPC that is controlled by deep reinforcement learning (DRL). A value-based deep Q-network (DQN) directly selects switching actions and jointly regulates the partial-voltage and output current. The control problem is formulated as a discrete-time Markov decision process, and a two-stage transfer learning scheme ensures safe, efficient deployment. Firstly, the DQN agent is trained in a high-fidelity simulation and then fine-tuned with a small set of experimental data to capture parasitic and modeling errors. The controller is integrated into a constant-current–constant-voltage (CC-CV) charging algorithm and validated over a full charging cycle of a 60 kWh EV battery. The proposed control scheme exhibits a settling time of approximately 2 ms in response to current reference variations while maintaining steady-state errors below 2% in current regulation and below 1% in partial voltage regulation. Simulation results show that the proposed DRL controller has a small steady-state tracking error and improved robustness to reference changes compared with conventional PI and sliding mode controllers. The low computational cost of the trained DQN policy also enables real-time execution on embedded platforms for EV charging. Full article

To meet the growing demand for autonomous decision-making and real-time optimization in industrial manufacturing, integrating Artificial Intelligence-Generated Content (AIGC) services with Industry 5.0 can enable real-time industrial intelligence. The effectiveness of a generative model is closely related to the current state of the production environment. However, existing studies often ignore the dynamic temporal relationship between generative models and production environments, especially in industrial scenarios with large model transmission delays and random AIGC task arrivals. Therefore, we define a novel metric, namely the Age of Model (AoM), to measure the freshness of generative models with respect to current industrial tasks. We then formulate an average-AoM-minimization problem that jointly considers LoRA-based fine-tuning, wireless transmission and resource allocation. To solve this problem, we propose a Hybrid-Action Multi-Agent Proximal Policy Optimization (HA-MAPPO) algorithm. The proposed algorithm follows the centralized training and decentralized execution (CTDE) paradigm and introduces a Main-Agent Priority State Strategy to support coordinated training and independent execution. In addition, a multi-head output structure is designed to handle the hybrid-action space, which includes discrete fine-tuning association decisions and continuous transmission resource allocation. Simulation results show that the proposed scheme outperforms all benchmark methods. Specifically, the cumulative rewards are improved by approximately $11.13\%$, $20.32\%$, $36.61\%$, and $38.78\%$ compared with the four benchmark algorithms, respectively. These results demonstrate that the proposed scheme can significantly reduce the average AoM while providing high-quality and timely industrial AIGC services. Full article

Maintaining high power delivery under uncertain landing positions is a key challenge for wireless charging of unmanned aerial vehicles (UAVs). This paper presents a data-driven power optimization method based on the Soft Actor-Critic algorithm for multi-transmitter single-receiver wireless power transfer (MTSR-WPT) systems. To support effective learning without explicit online parameter identification, a physics-informed dual-current state representation is constructed from measurable current responses, combining a zero-phase current with the current response under the applied phase command. The agent is trained using a reward defined directly from normalized load power, and the transmitter voltage phases serve as the control actions. In simulations of a five-transmitter system, the learned policy achieves about 97% of the theoretical maximum power in the training region and about 96% in the expanded evaluation region. Additional robustness studies show strong performance under moderate measurement noise and substantial recovery under model mismatch after short fine-tuning. Experimental validation on a physical prototype confirms the effectiveness of the method, yielding an average power improvement of 188% from a zero-phase baseline and reaching 87% of the maximum power measured on the hardware platform. These results support the proposed method as a practical data-driven alternative to model-dependent MTSR-WPT power optimization for UAV wireless charging. Full article

To address the efficiency and cost limitations of traditional manual cataloging, this study proposes a large language model-driven automated cataloging workflow in which the Metadata Extraction Agent (MEA), Description Cataloging Agent (DCA), Subject Analysis & Indexing Agent (SAIA), and Quality Control Agent (QCA) collaborate to perform cataloging tasks. Experiments are conducted using a dataset of over 33,000 CNMARC bibliographic records from a University Library, together with data from the Chinese Library Classification (5th edition). Meanwhile, the agent-based workflow framework directly employs large language models without additional enhancement techniques, thereby providing a useful experimental benchmark for evaluating future AI-assisted cataloging systems. The results show that the framework performs well in metadata recognition, bibliographic description, and macro-level classification tasks, and can relatively stably generate standardized records. However, limitations remain in fine-grained semantic indexing and the interpretation of complex contexts. Therefore, in light of the capability limitations revealed by the experimental results, the study argues that fully automated end-to-end cataloging relying solely on generative AI is not yet entirely feasible. Future improvements should integrate techniques such as retrieval-augmented generation, supervised fine-tuning, and structured reasoning prompts, while establishing traceable mechanisms to enhance the reliability of intelligent cataloging. Full article

Flood risk maps play a critical role in land-use regulation, infrastructure planning, and community preparedness, which are key components of sustainable and climate-resilient urban development. Their production, however, remains costly, labor-intensive, and time-demanding as it relies on simulation-driven workflows that combine hydrodynamic modeling with expert interpretation and extensive validation. To address this issue from a sustainability perspective, we develop a novel, practical, and near-real-time large language model (LLM)-based framework to support property-level flood risk assessment. This framework, which synthesizes geospatial, hydrological, infrastructural, and historical flood information, extends existing research and explores novel risk estimation methods for use in planning practice. Using Brisbane, Australia, as a case study, we develop Flood-LLM, a multi-agent system that transforms multi-source urban datasets into structured textual representations, models diverse neighborhood conditions, and fine-tunes a reasoning model using expert-assessed risk classifications. The results show that Flood-LLM can reproduce official flood risk labels for creek, river, storm tide, and overland-flow hazards with reasonable accuracy, outperforming classical machine learning, deep learning, and untuned LLM baselines. Visual and quantitative analyses indicate that the framework demonstrates a qualitatively nuanced capability to capture salient spatial patterns present in the official maps, while generating a textual chain-of-thought providing a transparent audit trail for its labeling decisions. These findings suggest that such LLM-based approaches can produce potential complementary tools to expert-reviewed planning classifications and support more sustainable, adaptive flood risk management by enabling timely map production and updates that facilitate informed decision-making in rapidly changing environmental conditions. Full article

Deploying Large Language Models (LLMs) on Maritime Autonomous Surface Ships (MASS) entails a critical trade-off between reasoning depth, inference latency, and hardware constraints. To fill the existing gap, we introduce MARTIAN (Maritime Agent for Real-time Tactical Inference And Navigation), a 14B-parameter decision support agent engineered for edge deployment on standard vessel hardware (e.g., the NVIDIA Jetson AGX Orin). Central to our approach is the Cognitive Core architecture, which utilizes a verified dataset of 21,800 Chain-of-Thought (CoT) instruction–response pairs to align general linguistic capabilities with maritime procedural logic. Empirical evaluations demonstrate that MARTIAN achieves an overall accuracy of 73.23% (SFT only) and 81.16% (SFT + RAG) on the Bilingual Maritime Multiple-Choice Questionnaire (BM-MCQ), a standardized assessment dataset constructed based on Officer of the Watch (OOW) competencies. Notably, the SFT-only configuration attains 78.53% on pure-logic-intensive COLREG tasks—surpassing the 72B-parameter Qwen-2.5 foundation model in this domain—while maintaining a real-time inference latency of 22.4 ms/token. Crucially, our ablation studies support a nuanced Interference Hypothesis: while RAG significantly enhances factual recall in knowledge-intensive domains (boosting total accuracy from 73.23% to 81.16%), it concurrently introduces semantic noise that degrades performance in pure logic reasoning tasks (e.g., COLREG maneuvering accuracy decreases from 78.53% to 77.36%). On the basis of this finding, we identify and empirically motivate a decoupled cognitive design principle that separates procedural reflexes (via SFT) from declarative knowledge (via RAG). While the full implementation of an adaptive routing mechanism is deferred to future work, the ablation results presented herein offer a validated, cost-effective reference architecture for deploying transparent and regulation-compliant AI on resource-constrained merchant vessels. Full article

A novel model-free hierarchical reinforcement learning (HRL)–based Linear Quadratic Regulator (LQR) control framework with adaptive weight selection is proposed to address the reliance of conventional LQR methods on accurate system models and manual parameter tuning. The proposed approach adopts a two-level learning architecture in which a high-level meta-agent adaptively optimizes the LQR weighting matrices Q and R through entropy-based trajectory evaluation, while a low-level base-agent performs model-free policy iteration to update the state-feedback control law under unknown system dynamics. By decoupling weight optimization from control-law learning, the framework enables simultaneous adaptation of the cost-function parameters and the feedback gain without requiring explicit model information. To enhance learning stability and exploration during weight adaptation, Gaussian noise and an experience replay mechanism are incorporated into the learning process. Numerical simulations on second- and third-order linear systems demonstrate that the proposed HRL-based LQR method achieves effective control performance, reliable convergence, and improved adaptability in model-free environments. Full article

The optimization of the MOSFET gate drive parameters is crucial for the trade-off between switching loss and electromagnetic interference (EMI). However, the nonlinear coupling among gate drive parameters, board-level parasitic, and switching performance limits the effectiveness of traditional MOSFET drive design methods. This paper proposes an adaptive tuning framework based on the proximal policy optimization (PPO) algorithm. An analytical switching model incorporating board-level parasitics is first derived to analyze the coupling between drive parameters and switching performance. The optimization problem is then formulated as a Markov decision process (MDP). Within this framework, domain randomization is applied during training. This enables the agent to learn a generalizable optimization strategy that remains robust across the varying parasitic inductances encountered in different PCB layouts. Compared to the traditional Non-dominated Sorting Genetic Algorithm II (NSGA-II), the proposed method uses the trained policy for direct inference. This reduces computation time by $98.7\%$ while maintaining a multi-objective performance difference within $10.06\%$. In addition, hardware verification shows a 10.7% average deviation between the measured and simulated results. These results demonstrate that the proposed method provides an efficient and scalable solution for MOSFET gate drive optimization. Full article

In GNSS-denied large-scale outdoor environments, UAVs and UGVs that rely solely on visual–inertial odometry (VIO) suffer from accumulated global drift as the trajectory grows. Meanwhile, inter-platform ultra-wideband (UWB) ranging exhibits unknown, time-varying noise under NLOS/multipath, rendering naïve weighting unreliable. This paper presents an anchor-free collaborative localization framework for UAV–UGV teams that fuses pairwise UWB ranges (including UAV–UAV, UAV–UGV, and UGV–UGV) with onboard VIO in a factor-graph backend via a two-stage robust scheme. First, we bound VIO drift using per-agent state covariance and reject UWB outliers with a Mahalanobis gate, preventing early-stage bias when VIO is still accurate. Then, during global optimization, we adaptively estimate the Fisher information of UWB factors from measurement–state residuals, enabling online self-tuning of measurement confidence under time-varying SNR. Real-world experiments with three UAVs and two UGVs over multi-level rooftops and forest–open areas (~1.6 km²) show that, compared to an outlier-only variant, the proposed method further reduces localization RMSE by about 24.6% and maximum error by about 31.2% for both UAVs and UGVs, maintaining strong performance during long trajectories dominated by VIO drift and NLOS ranges. The approach requires no fixed anchors or GNSS and is applicable to UAV–UGV teams for disaster response, cooperative mapping/inspection, and bandwidth-limited operations. Full article

The high-speed Electromagnetic Suspension (EMS) maglev guidance system exhibits inherent characteristics of strong nonlinearity, parameter time-variation, and complex external disturbances. To further optimize and improve the control performance of the guidance system for high-speed maglev trains, a novel intelligent control strategy that integrates the Deep Deterministic Policy Gradient (DDPG) algorithm with Super-Twisting Sliding Mode Control (STSMC) is proposed. Focusing on a single-ended guidance unit with differential control of dual electromagnets, an STSMC controller is first designed based on a cascaded control framework. To overcome the limitation of offline parameter tuning in dynamic operational conditions, a reinforcement learning optimization framework employing DDPG is introduced. A multi-objective hybrid reward function is formulated, incorporating error convergence, sliding mode stability, and chattering suppression, thereby realizing the online self-tuning of core STSMC parameters via real-time interaction between the agent and the environment. Numerical simulations under typical disturbance conditions verify that the proposed DDPG-STSMC controller significantly reduces the amplitude of guidance gap variation and accelerates dynamic recovery compared to conventional PID control. Its superior performance in disturbance rejection, control accuracy, and operational adaptability is validated. This study, conducted through high-fidelity numerical simulations based on actual system parameters, provides a robust theoretical foundation for subsequent hardware-in-the-loop (HIL) experimentation. Full article

Drones are expected to support applications such as emergency response, parcel delivery, and infrastructure monitoring in dense urban airspaces, creating traffic levels that are unmanageable for human operators. Autonomous separation management is therefore essential, combining strategic and tactical control to prevent conflicts. This paper addresses the tactical landing phase by introducing a centralized landing flow manager—a reinforcement learning (RL) agent that adjusts drone speed and heading to merge landing flows safely and efficiently prior to a final approach fix. The objective of the work was to demonstrate the potential of reinforcement learning in this novel context, by implementing and evaluating it in simulation and testing its capabilities with 10 concurrent landing drones. The RL agent learns to successfully separate traffic, thereby lowering intrusion counts compared to the baseline autopilot, but is outperformed in safety by the decentralized Modified Voltage Potential (MVP) method due to outlier scenarios. Nevertheless, the RL-based system achieves faster scenario completion and thus a higher overall throughput, by speeding up the vehicles towards the final approach fix. Future work will explore improved network architectures, transfer learning across varied scenarios, and algorithmic fine-tuning to further enhance safety performance. Full article

One-dimensional alumina nanorods have garnered significant attention due to their unique physical and chemical properties, which hold great promise for applications in catalysis, sensing, and other fields. However, the precise control over the morphology and properties of these nanorods remains a challenge, particularly in achieving a high specific surface area and desirable crystallinity. In this work, we explored the hydrothermal synthesis of alumina nanorods, focusing on the effects of structure-directing agents. It was observed that extending the hydrothermal time and optimizing the temperature led to the formation of nanorods with enhanced crystallinity and specific surface area. The addition of urea and different structure-directing agents significantly influenced the morphology and properties of the nanorods. Furthermore, density functional theory (DFT) calculations revealed the underlying mechanisms of how these structure-directing agents affect the adsorption and growth of alumina nanorods on different crystal planes. Our findings suggest that by carefully tuning these parameters, it is possible to achieve alumina nanorods with optimized properties. This work not only provides a systematic approach to the synthesis of alumina nanorods but also opens up new possibilities for the development of advanced materials with tailored properties for a wide range of applications. Full article