Search Results (766)

Despite the large amount of successful existing methods and frameworks for planning sets of multiple unmanned aerial systems (UASs), there is still a lack of coordination frameworks that are capable of coping with real-world operational conditions. This paper presents U-Plan, an integrated management framework for the coordination of multi-UAS missions. U-Plan is designed to plan, schedule, monitor, and replan a heterogeneous set of UASs to complete point of interest (PoI) visiting missions while ensuring that all the generated trajectories are safe, feasible, and compliant with the required PoIs’ arrival times, UAS kinematics and energy constraints, and the existing 3D no-fly zones (NFZs). U-Plan is designed as a practical tool for strongly dynamic missions and is built upon three core components: (1) an NFZ-aware route computation method that explicitly accounts for NFZs prior to vehicle routing problem (VRP) optimization, resulting in shorter NFZ-safe routes; (2) a trajectory smoothing module that ensures the generation of kinematically feasible trajectories for fixed-wing UASs; and (3) a mission supervision module for real-time monitoring and replanning in case of changes in the UAS, mission, wind speed, or airspace restrictions. To validate the proposed architecture, we conducted rigorous experiments utilizing the VECTOR-SIL autopilot and Visionair Ground Control Station to realistically replicate the behavior of certified fixed-wing autopilots under various weather conditions using the exact same hardware and flight control software that runs onboard the physical drones. The validation shows U-Plan’s capacity to efficiently satisfy complex mission requirements with strong scalability. Due to its high computational efficiency, U-Plan enables online mission replanning, allowing UAS fleets to seamlessly adapt to changes that are typical of real-world operational scenarios. Full article

Combinatorial optimization is a key component in critical decision problems such as routing, scheduling, network design, and graph optimization. Although combinatorial optimization methods, including exact algorithms, approximation methods, constraint programming, mixed integer programming, and metaheuristics, are widely available, they often face obstacles, such as limited scalability and adaptability in various applications. In this study, a systematic critical review of machine learning for combinatorial optimization is provided to characterize the usage and evaluation of learning-based approaches. A detailed analysis is used to infer and determine findings and limitations. The paper emphasizes how machine learning for computational optimization has changed over time, moving from end-to-end neural solvers to hybrid systems. Learning components are essential for directing, speeding up, or enhancing traditional solver backbones such as constraint programming and metaheuristics in hybrid systems. The review also critically examines current limits that impact performance in general, including scalability, deployment readiness, generalization, and benchmark consistency. Even though using large language models for problem formulation and heuristic synthesis has potential, more work needs to be done to ensure reliable validation. As a conclusion, this article examines recent studies’ findings, emphasizes the growing trend toward hybrid learning-driven optimization frameworks, and underlines important methodological limits and unresolved issues. Full article

To scientifically explore the vehicle capacity characteristics of circular earthwork transportation routes in water conservancy projects, this paper takes the second-phase project of the Huaihe River Sea Entrance Channel as the research background. Key influencing factors such as road conditions, vehicle performance parameters, safe car-following distance, and earthwork loading–unloading duration are comprehensively considered, and a cellular automaton simulation model is constructed. Horizontal comparative verification is carried out with the Intelligent Driver Model, System Dynamics model, and field measured data to verify model accuracy. The results reveal that the cellular automaton (CA) model yields a total vehicle transport trip count of 606, with a MAPE of 0.66% when compared against the field-measured average of 602 trips. The simulated average travel speed reaches 16.71 km/h, corresponding to a MAPE of 2.89% relative to the field measurement of 16.24 km/h. The error metrics of these two indicators are markedly lower than those derived from alternative models. Due to differences in modeling paradigms and applicable mechanisms, the three models exhibit distinct characteristics in simulation performance. Among them, the cellular automaton model is more suitable for the circular earthwork transportation scenario of this study, which can accurately reflect the coupling characteristics of microscopic traffic behaviors such as multi-route confluence and node queuing, and has high consistency with actual engineering operation. Sensitivity analysis indicates that improving earth loading efficiency and reasonably arranging excavator quantity can significantly enhance the overall transportation efficiency. The modeling ideas and simulation analysis method adopted in this paper are not only applicable to the specific engineering scenario, but also can be extended to similar water conservancy earthwork transportation and large-scale engineering logistics transportation fields. It can provide theoretical basis and engineering reference for earthwork scheduling optimization and quantitative calculation of traffic capacity in water conservancy projects. Full article

Long-context large language models incur substantial computational overhead during autoregressive decoding. Existing sparse attention methods can improve inference efficiency, but they typically rely on fixed sparse patterns, historical attention statistics, or coarse-grained proxy representations to estimate important KV positions, making it difficult to accurately capture query-dependent fine-grained relevance for dynamic KV retrieval. In this paper, we propose Block-then-Hash Attention (BTHA), a two-stage KV retrieval method: it first performs block-level routing with mean key representations to rapidly reduce the candidate search space, and then applies a learnable orthogonal hash network within the routed KV candidates for fine-grained token-level position retrieval. The hash network is trained offline to learn the hash mapping between queries and keys, with a low training cost: on Llama-3.1-8B-Instruct, training can be completed in approximately two hours using a single NVIDIA A100 GPU. During inference, BTHA implements block-level routing, hash-based retrieval, and sparse attention computation with dedicated operators, and further employs CPU–GPU collaborative scheduling to reduce memory access, synchronization, and candidate selection overhead, thereby achieving end-to-end decoding acceleration. Extensive experiments on LongBench-E show that BTHA consistently outperforms state-of-the-art top-K attention methods in both accuracy and efficiency; under a 512-position budget, it achieves the best average accuracy on both Llama-2-7B-32K-Instruct and Llama-3.1-8B-Instruct, while delivering up to 7.0× speedup over vanilla full attention. Full article

The Annual Delivery Program (ADP) for Liquefied Natural Gas (LNG) represents a complex maritime inventory-routing problem that requires the precise synchronization of production and distribution. This study introduces a novel Mixed Integer Linear Programming (MILP) model designed to optimize vessel routing and scheduling over a one-year horizon under a direct-shipment assumption. The model minimizes total logistics costs, encompassing both fixed annual fleet costs and daily operating costs. The novelty of the model can be summarized in two aspects. First, it simultaneously optimizes several decisions: the assignment of frequency of deliveries to customers, the assignment of vessels to customers, cargo load sizes, and vessel routing and scheduling. The key distinction is that, unlike existing formulations that take the frequency of deliveries to customers as a fixed parameter, this frequency is itself a decision variable selected from a customer-specific discrete set; the selected frequency partitions the planning horizon into uniform windows and sets each delivery’s cargo load size to the exact demand accumulated over its window from daily demand data. Second, it incorporates several relaxations of selected variables and valid inequalities that enable us to solve the complex model for moderate size problems within a reasonable computational time using the exact optimization approach. Using this novel model, we carried out extensive numerical analysis based on cost and operational parameter scenarios and developed important insights for the characteristics of a solution to the problem. Full article

Urban Unmanned Aerial Vehicle (UAV) logistics operations are frequently constrained by the intersection of limited battery endurance and dynamic ground traffic. When mobile depots are delayed by congestion, onboard drone fleets experience extended idling periods, leading to constrained sortie generation and reduced asset utilization. To address this bottleneck, this paper introduces a traffic-predictive multi-UAV dispatch framework for deterministic day-ahead planning under modeled urban operating conditions. By coupling a count-derived macroscopic speed surrogate learned using XGBoost with a Particle Swarm Optimization (PSO)–Mixed-Integer Linear Programming (MILP) optimization architecture, the framework synchronizes mobile depot trajectories with forecasted low-congestion windows and pre-allocates endurance-feasible parallel aerial sorties. Controlled computational experiments across 30 synthetic routing instances demonstrate the potential value of this approach within the stated modeling assumptions. Compared to baseline clustered deployments, the traffic-aware framework raises mean fleet utilization from 0.43 to 0.63—a 46.2% relative improvement driven by temporal compression of the mission window rather than an absolute increase in flight hours. Furthermore, the proposed framework reduces total mission completion time by 69.87% relative to the conventional truck-only baseline, while achieving a 29.58% incremental gain over static speed drone deployments. These findings suggest that incorporating predictive ground traffic information into day-ahead UAV scheduling can improve modeled fleet efficiency; however, field validation with measured route-level speeds, real delivery demand, and operational constraints remains necessary before deployment-level claims can be made. Full article

Discrete diffusion models have become an important class of generative models for categorical data, yet their theoretical understanding remains largely limited to time-homogeneous noise schedules. In this work, we study uniform-rate discrete diffusion models with time-inhomogeneous continuous-time Markov chain forward processes. We establish convergence guarantees for practical reverse-time samplers by directly controlling the total variation distance, avoiding the indirect route of first bounding KL divergence and then applying Pinsker’s inequality. Our analysis decomposes the sampling error into initialization, score-estimation, discretization, and early-stopping errors, and explicitly characterizes how each term depends on the accumulated noise, the local noise rate, and the smoothness of the noise schedule. Under suitable regularity conditions on the noise schedule, we further derive step-complexity guarantees that match the order of existing results for homogeneous samplers. Full article

Static pre-scheduling in inland container depot (ICD)-centered drayage must coordinate tractors, detachable load units, factory loading, and port deadlines before next-day execution. Conventional order-based routing is too rigid for mixed direct haulage, drop-and-pull, relay pickup, street-turn, and buffering operations. This study proposes a task-pooling framework that decomposes logistics orders into atomic tasks and recombines them across tractors in a unified static planning space. A compact route-based MILP is used for reduced-scale calibration, and an enhanced adaptive large neighborhood search (E-ALNS) is developed around ICD-oriented relay recombination and temporal-slack shifting. On a realistic synthetic benchmark with 100 generated order records (90 active executable orders), 60 available tractors, and 330 executable tasks, the proposed method reduces the internal search-ledger value from 42,213.29 to 34,421.22 and the compact ex post blueprint value from 53,802.28 to 47,717.99 relative to the greedy construction baseline. The resulting blueprint preserves an average inter-task slack of 89.86 min and a 5th-percentile slack of 61.73 min. A generic adaptive-neighborhood baseline reaches a slightly lower ex post value of 46,722.48 only with a longer runtime and much lower temporal reserve. The results support a cost–reserve–runtime tradeoff interpretation rather than unconditional cost dominance. Full article

Multi-runway sequencing of unmanned aerial vehicles (UAVs) at temporary disaster relief aerodromes presents a priority-heterogeneous scheduling problem under class-asymmetric wake turbulence constraints. We formulate this as a priority-weighted Markov decision process with a deliberately minimalist reward—per-step class weights for completed landings, with no shaping or hand-crafted safety logic—and extend it with per-UAV operational deadlines (encoding en-route endurance consumption) and per-runway queue capacity constraints that produce a non-trivial action mask. We train a Proximal Policy Optimisation (PPO) agent and benchmark it against six baselines spanning deterministic optimisation (Joint-LA-1), stochastic lookahead (Stochastic-LA), and online tree search (MCTS). Across 100 paired evaluation episodes, PPO matches the operational standard Priority-FCFS within 2.7% (p = 0.124, not significant); Joint-LA-1, the strongest non-learned baseline, outperforms PPO by 3.2% (p = 0.043). Despite near-identical aggregate throughput, PPO autonomously develops a runway specialisation pattern—concentrating 60% of high-priority landings on a single strip while routing 93% of emergency arrivals to the remaining strips—that emerges entirely from the reward signal. Under looser deadlines, the PPO–PFCFS gap narrows to −0.5%, and wake symmetry ablation reveals that PPO outperforms Priority-FCFS by 46.5% when the asymmetric wake structure is removed. These results demonstrate that priority-aware capacity reservation can emerge without embedded domain knowledge, and that simple heuristics are near-optimal under tight operational constraints—a finding with direct implications for autonomous scheduling in disaster relief aviation. Full article

With more satellites, richer payload resources, and more diverse service functions, satellite systems are increasingly operated as large space–ground networks. These networks must schedule arriving missions under changing topology, gateway access, beam availability, weather-affected links, spectrum compatibility, and mission time windows. Offline optimization can compute high-quality schedules when the mission set, satellite visibility windows, and resource states are known before execution, but repeated replanning is costly for asynchronous arrivals. Online heuristics make faster decisions from local route rules, but they do not evaluate how an accepted service path changes the capacity left for later requests. Reinforcement-learning schedulers can adapt from delayed scheduling outcomes. However, many generic policies rely on fixed-step state updates or flat compound-action scores, whereas online satellite scheduling makes decisions at irregular arrivals over continuously evolving topology and capacity-coupled service paths. We propose TopoAgent, an online reinforcement-learning agent for heterogeneous satellite mission scheduling. TopoAgent models each request as a service-path decision, propagates compound feasibility through the satellite–gateway–beam hierarchy, and uses a capacity-aware policy to choose among feasible paths. A deterministic constraint manager places the selected path in time, while SRV guides the policy toward assignments that preserve reusable beam capacity. In a high-fidelity simulator, TopoAgent achieves a 74.7% mission completion rate and a 75.5% high-priority completion ratio over five seeds. Full article

The continuous growth of IIoT systems has significantly increased the number of connected devices and message interactions, creating higher requirements for communication mechanisms in terms of scalability and adaptability under dynamic network environments. Although MQTT is widely used for its lightweight communication, its traditional centralized broker architecture limits scalability and fault tolerance in large-scale data transmission, reducing system scalability and fault tolerance. Additionally, static QoS configuration is difficult to adapt to dynamic environmental changes, resulting in high end-to-end latency and limited system throughput. To address these issues, this paper proposes a distributed MQTT communication method based on intelligent QoS routing and hierarchical collaboration (DAIS-MQTT). This method designs a network routing algorithm based on a hierarchical tree structure (LCN), which effectively addresses the scalability limitation of centralized proxies by enabling multi-level proxy collaboration and self-recovery from faults. At the same time, it proposes a QoS routing algorithm based on intelligent decision trees (IQR), which jointly optimizes proxy selection and QoS levels to dynamically adapt to changes in the network environment, thereby solving the problem of insufficient adaptability in static QoS configurations. Experimental results show that compared with the traditional MQTT-based communication method, the DAIS-MQTT method reduces the average message delay by 29.9%, increases system throughput by 28.2%, and maintains a reliable transmission rate of 98.7% in unreliable network environments, making it suitable for high-dynamic and large-scale IIoT communication scenarios. Full article

A Software-Defined Network (SDN) renders flexible traffic engineering, but consumes a lot of energy. There is an overhead on the control-plane because flow rule updates are always performed and there is energy consumption by the forwarding hardware. Current energy-aware SDN methods mostly focus on Static or Greedy optimizations. This can cause too many Ternary Content-Addressable Memory (TCAM) updates and unstable rule churn when traffic changes over time. This article introduces a Dynamic Flow Rule Placement (DFRP) framework for real-time energy optimization in SDN. It reduces network energy usage, TCAM update costs, and rule churn all at the same time. The suggested framework uses a convex relaxation method to take decisions on binary switches, links, and rule placement. It also uses a minimum-edit round scheme that only allows small rule changes between time slots. To further reduce instability in the control plane, batch scheduling and receding horizon optimization (RHO) techniques are used. The system uses predicted traffic for future time slots to make decisions, but only the actions for the current time slot are executed. The experiments are carried out on two real-world dynamic SNDlib topologies such as Germany50 and Nobel-Germany, using 288 five-minute traffic matrices over a one-day period. Comparative results against Static and Greedy baselines show that DFRP saves approx. 30% energy while cutting down on TCAM update overhead and rule churn by approx. 20%, consistently across both the networks. Hence DFRP can be applied on dynamic traffic large-scale networks for stable and energy-efficient SDN operations. Full article

Background/Objectives: Pharmaceutical formulation involves designing a drug product by combining the properties of an active pharmaceutical ingredient (API) with suitable excipients and processing strategies to produce a safe, effective, and manufacturable dosage form. However, data in formulation science are often limited, expensive to generate, and frequently restricted by proprietary and confidentiality constraints. Interactive digital tools can support formulators during early drug product development by improving the structure, transparency, and efficiency of formulation decision-making. While the current system focuses on structured decision support, future extensions may incorporate machine-learning methods for recommendation and knowledge extraction. Methods: In this work, we developed the Formulation tool, an interactive decision-support and visualization system for formulation development based on a hierarchical formulation-strategy framework commonly used in pharmaceutical practice. The tool is designed to prioritize suitable formulation principles and associated processing routes, with oral solid formulation as the initial application domain. The evaluated scenarios also include pathway regions relevant to oral liquid formulations. Its modular architecture also makes it adaptable to other formulation scenarios. To assess practical applicability, the tool was evaluated in a structured expert study involving five expert users across six predefined formulation scenarios (n = 30 runs), covering three drugs under adult and pediatric conditions. Results: The tool showed agreement with the expected dosage-form families and overall formulation properties, with adult scenarios converging to oral solid regions and pediatric scenarios converging to oral liquid regions. At the downstream formulation-profile level, users converged either to the dominant expected pathway or to alternative feasible pathways within the same formulation region. Variability in full pathway completion was observed and was primarily associated with differences in user interaction behavior and exploratory usage patterns. The median task completion time was 113.5 s. Conclusions: In addition to organizing formulation knowledge, the Formulation tool records user interactions in a structured manner, which may support future learning from usage patterns. Because detailed downstream formulation constraints are often institution-specific and are typically not available in the public domain, the present evaluation focused on agreement at the dosage-form-family level and on overall formulation properties rather than on highly specialized constraint logic. The system is based on a constraint satisfaction problem (CSP) framework, which is well suited for modeling complex decision processes under explicit constraints. CSP has also been widely applied in intelligent scheduling systems, supporting its suitability for structured, constraint-rich decision-making tasks such as pharmaceutical formulation strategy development. Full article

Large-scale WebGIS visualization of 3D building models is often constrained by large requested payloads, client-side memory pressure, and runtime state-parsing overhead. This study proposes a five-dimensional orthogonal decoupling framework and a 16-bit state-word-driven scheduling method for 3D building models. The Boundary-based Spatial Proxy–Geometric Detail–Component Complexity–Texture Appearance–Semantic Information (B-D-C-T-S) framework organizes model representations into five separately addressable and schedulable dimensions, covering spatial proxies, geometry, components, textures, and semantics. A compact 16-bit structured state word is used to represent runtime states and reduce dependence on repeated text-based state parsing, supporting fixed-offset bitwise decoding, exclusive-OR (XOR)-based differencing, constraint checking, and incremental updating. A centroid-assigned Home Tile strategy is further introduced to reduce redundant semantic payloads for cross-tile objects. The method was evaluated using a single-building BIM model and an urban-scale photogrammetric mesh dataset. Under the tested initial-view setting, staged decoupled loading reduced the first-screen requested payload by 93.1% compared with monolithic loading. State-word-based C-field extraction achieved an approximately 144-fold speedup over JSON deserialization and C-field lookup. The Home Tile strategy reduced the total semantic payload by 44.1% in the semantic-redundancy test. In the 1.12 GB first-screen memory test, state-word-driven D1 tile scheduling loaded only 22.7 MB of physical payload, with stable resident memory of approximately 88.1 MB. These results indicate that the proposed method supports object-level state representation, selective resource activation and scheduling, Home Tile semantic routing, incremental updating, and first-screen memory control within tiled Web3D pipelines. Full article

Because high-precision DEMs are costly to obtain, while low-precision DEMs often fail to meet accuracy requirements for watershed flood simulation, this study proposes a multi-source DEM fusion method based on the Random Forest algorithm. This method combines K-Means slope clustering and Optuna hyperparameter optimization to realize adaptive weight allocation across eight slope zones. After multi-source DEM fusion, the fused DEM is applied to the flood simulation model of the Wei River Basin to simulate the catastrophic flood event in July 2021. The results show that the Mean Absolute Error (MAE) of the fused DEM ranges from 0.9855 to 1.7218, the Root Mean Square Error (RMSE) ranges from 1.0902 to 2.3953, and the Mean Error (ME) is close to 0 with no significant systematic bias. Compared with single-source DEM, the fused DEM reduces MAE by 21.32–85.32% and RMSE by 7.63–82.03%. In flood simulation, the peak discharge error based on the fused DEM is controlled within 0.013–0.059, and the coefficient of determination (R²) is not less than 0.9808. The simulated errors of inundation area and flood detention volume in flood detention areas are significantly lower than those using a single-source DEM. The proposed multi-source DEM fusion method can effectively improve terrain accuracy and the reliability of flood routing simulation, providing technical support for flood control scheduling in the Wei River Basin and watershed hydrological and flood simulation in data-scarce regions. Full article