Autonomous collision avoidance is a critical requirement for Maritime Autonomous Surface Ships (MASS), especially in restricted waterways with limited maneuvering space, dense traffic, and strict regulatory constraints. Existing model-based and reinforcement learning approaches often simplify the action or state space, resulting in poor maneuver feasibility and limited rule compliance in narrow channels and harbor environments. This paper proposes a POMDP-based collision avoidance framework tailored for restricted-water navigation. The proposed method integrates vessel maneuverability constraints, navigational boundary limitations, and rule-aware semantic representations to capture encounter situations and local regulatory priorities. A two-dimensional continuous action space, combining heading rate and speed adjustments, is introduced to improve maneuver realism and feasibility. The enhanced POMDP is trained using Proximal Policy Optimization (PPO), enabling stable policy learning under partial observability and dynamic multi-vessel interactions. Simulation results demonstrate that the proposed approach effectively ensures safe, efficient, and rule-compliant navigation in scenarios involving both static obstacles and dynamic vessels, validating the advantages of the improved state and action representations.

With the rapid expansion of waterway transport, accurate and efficient inland vessel draft detection has become a key component of navigational safety monitoring. Existing vessel draft detection methods have obvious shortcomings in cross-scenario generalization ability and robustness, especially in scenarios where the vessel hull is covered with stains or the draft scale characters are missing. These unwanted artifacts will lead to detection errors and missed detections, which makes it difficult to meet the needs of practical supervision. To address these issues, this paper proposes a high-precision detection method based on multi-task learning and weighted directed graphs, termed VDDNet (Vessel Draft Detection Network), aiming to achieve high-precision and realtime detection under complex scenarios such as hull stains and missing draft scale characters. First, the proposed method uses the YOLOv11 network, which retains anchor-free advantages and enhances feature capture for small or occluded targets via multi-scale fusion, to extract the draft scale region. On this basis, a multi-task learning architecture is constructed to jointly optimize draft scale character detection and vessel/water segmentation. In the vessel/water segmentation subtask, a channel and spatial attention module (CBAM) is further embedded in VDDNet to strengthen the extraction of vessel/water boundary features, thereby effectively reducing the impact of background interference on segmentation results. To eliminate the detection deviation caused by missing draft scale characters and stain interference, a weighted directed graph model is introduced to reconstruct the draft scale information. The accurate rearrangement of draft scale characters is then completed by combining geometric constraint rules. Finally, a dynamic reading calculation strategy is designed to achieve the accurate vessel draft detection results. Experimental results show that our VDDNet method can maintain high-accuracy and stable detection results under complex conditions such as hull stains and missing characters. Specifically, the proposed method achieves a mean absolute vertical distance (MAVD) of 2.006 pixels and a mean absolute draft depth error (MADDE) of 0.04 m. Meanwhile, the corresponding inference speed can reach 80 FPS, which can fully meet the real-time requirements of inland waterway traffic monitoring scenarios.

Aircraft assembly is characterized by stringent precedence constraints, limited resource availability, spatial restrictions, and a high degree of manual intervention. These factors lead to considerable variability in operator workloads and significantly increase the complexity of scheduling. To address this challenge, this study investigates the Aircraft Pulsating Assembly Line Scheduling Problem (APALSP) under skilled operator allocation, with the objective of minimizing assembly completion time. A mathematical model considering skilled operator allocation is developed, and a Q-Learning improved Particle Swarm Optimization algorithm (QLPSO) is proposed. In the algorithm design, a reverse scheduling strategy is adopted to effectively manage large-scale precedence constraints. Moreover, a reverse sequence encoding method is introduced to generate operation sequences, while a time decoding mechanism is employed to determine completion times. The problem is further reformulated as a Markov Decision Process (MDP) with explicitly defined state and action spaces. Within QLPSO, the Q-learning mechanism adaptively adjusts inertia weights and learning factors, thereby achieving a balance between exploration capability and convergence performance. To validate the effectiveness of the proposed approach, extensive computational experiments are conducted on benchmark instances of different scales, including small, medium, large, and ultra-large cases. The results demonstrate that QLPSO consistently delivers stable and high-quality solutions across all scenarios. In ultra-large-scale instances, it improves the best solution by 25.2% compared with the Genetic Algorithm (GA) and enhances the average solution by 16.9% over the Q-learning algorithm, showing clear advantages over the comparative methods. These findings not only confirm the effectiveness of the proposed algorithm but also provide valuable theoretical references and practical guidance for the intelligent scheduling optimization of aircraft pulsating assembly lines.

ObjectiveThis paper introduces a novel data-driven approach for generating realistic and hazardous overtaking scenarios. These scenarios are crucial for rigorously evaluating the autonomous collision avoidance capabilities of autonomous ships. Existing methods often struggle to balance scenario diversity, realism, and the representation of hazardous situations. To overcome this limitation, our method leverages the rich information embedded in automatic identification system (AIS) data to generate diverse and realistic overtaking encounters. MethodSpecifically, we propose a hybrid model that integrates a sequence generative adversarial network (SeqGAN) with a self-attention mechanism (SAM). The SeqGAN captures the complex patterns and dynamics in AIS-based ship trajectories, enabling the generation of novel, yet plausible, overtaking maneuvers. The incorporation of a SAM further enhances the model's ability to capture long-range dependencies in ship trajectories, resulting in more realistic and nuanced simulations. To ensure that the generated scenarios accurately reflect hazardous situations, we have developed a constraint model based on longitudinal and lateral safety distances between vessels to define realistic initial conditions. This model dynamically adjusts the initial positions and velocities of both the target vessel and the autonomous ship under test, ensuring that each generated scenario presents a genuine collision risk. ResultsThe results show that the effectiveness of our approach is validated through extensive simulations. A total of 500 hazardous overtaking scenarios were generated, significantly improving the coverage of test scenarios. Notably, 97.3% of these generated trajectories fall within a predefined buffer zone that encompasses real-world trajectories, demonstrating the high fidelity of our model. Furthermore, the speed distributions of the generated target vessels closely match those observed in real-world AIS data, further validating the realism of our approach. ConclusionThe enhanced realism and diversity of scenarios generated by this method significantly improve the efficiency of autonomous collision avoidance testing. This allows for a more precise definition of safety performance boundaries and accelerates the development and optimization of autonomous collision avoidance algorithms. Ultimately, this work contributes to the development of safer and more reliable autonomous maritime systems capable of navigating the complexities of modern maritime environments.

Long-horizon target navigation requires a robot to sustain task execution across evolving observations, decisions, and physical interactions. This requires three coupled capabilities: maintaining valid scene memory, revising target beliefs under partial observability, and selecting interaction-feasible navigation endpoints. However, the state underlying each capability is only conditionally valid: scene representations become stale when objects move or disappear, unsuccessful searches alter beliefs over target locations, and geometrically convenient endpoints may still be infeasible for manipulation. To address these challenges, we present OmniNav, which formulates long-horizon navigation as continual inference over a factorized task state posterior coupling scene validity, target belief, and interaction feasibility. For representation, OmniNav incrementally constructs an updatable 3D object scene memory, preventing stale scene evidence from propagating to subsequent decisions. For exploration, it introduces an evidence-aware Bayesian belief-revision mechanism that derives dependency-aware region priors from semantic context, incorporates unsuccessful searches as negative evidence, and updates them for posterior-guided frontier selection. For interaction, OmniNav incorporates manipulation reachability and collision constraints into navigation-endpoint selection and propagates execution feedback through hierarchical closed-loop recovery. Extensive experiments demonstrate that OmniNav achieves the highest success rates among the compared methods on semantic ObjectNav and fine-grained instance navigation benchmarks, remains robust to target relocation, and improves real-world pick-and-place success from 53.3

ObjectiveTo address the challenge of ineffective communication regarding avoidance intentions between autonomous ships （AS） and traditional manned ships (hereafter referred to as TS) in a mixed traffic environment, this study proposes an interactive collision avoidance decision-making method based on Stackelberg game (S-G) theory and Chain of Thought (COT). The aim is to enhance the interactive collision avoidance decision-making capabilities of ships operating in mixed environments.Method First, ship collision avoidance scenarios in mixed environments are defined, and relevant research hypotheses are proposed. AS and TS ships are modeled using a leader-follower S-G game framework, with strategy spaces and payoff functions designed from a navigational practice perspective. Next, considering the interaction process between ships, a COT-game collision avoidance (COT-GCA) algorithm is developed, consisting of four sub-modules: state perception, intention sharing, strategy negotiation, and collision avoidance decision-making. Finally, the effectiveness of the proposed method is verified through experiments involving three-ship and four-ship encounter situations.ResultsThe experimental results demonstrate that ships in both groups can efficiently understand the avoidance intentions of other ships and successfully avoid collisions. The response time, steering range, and resumption of collision avoidance behavior exhibit timeliness, precision, and stability. The average output efficiency evaluation values before and after decision-making, calculated using the decision unit evaluation method, are 1 and 0.993, respectively, indicating the high efficiency of the S-G model in solving ship interaction collision avoidance problems.ConclusionThe proposed model and algorithm effectively enhance the interactive collision avoidance decision-making capabilities of ships in mixed environments, providing significant theoretical insights for future practical applications.

On-policy self-distillation (OPSD) adapts a language model by distilling guidance from a frozen teacher on trajectories sampled from the student. Its effectiveness, however, depends critically on the quality of those trajectories. We show that when student rollouts drift from target trajectories, conditioning the teacher on off-target prefixes substantially weakens its task-relevant supervision. Controlled prefix-corruption experiments expose this failure mode, which we term rollout-conditioned signal degradation. To address this problem, we propose a unified training framework that separates two complementary supervision pathways. The first retains rollout-conditioned distribution matching, providing guidance on states the student actually visits. The second applies supervised cross-entropy on canonical ground-truth contexts, avoiding the incompatibility of imposing target tokens on erroneous rollout prefixes. Token-level rollout-target alignment is used to adapt the strength of the canonical-context anchor, emphasizing it during cold start and relaxing it as rollout quality improves. Experiments across multiple model scales, two task families, and general-reasoning benchmarks show that the proposed approach improves task acquisition over OPSD while preserving general capabilities, resulting in a more favorable empirical plasticity-stability trade-off. These findings identify context quality as a central bottleneck in on-policy self-distillation and demonstrate the value of separating rollout-conditioned guidance from canonical supervision.

Offshore wind power is crucial for achieving carbon neutrality in East Asia and North America. However, these regions are prone to tropical cyclones, raising concerns about their resilience against such persistent threats. While previous work has focused on turbine fragility and single-event resilience, system-level wind farm resilience under multi-decadal climate considering cyclone-induced damage and maintenance interruption, stochastic component degradation and actual repair constraints remains largely unexplored. Here, we investigate multi-decadal offshore wind farm resilience under these interrelated factors using a proposed component-level, full-year, hourly failure-and-repair simulation model, and adopting reliable tropical cyclone, oceanographic, wind farm, maintenance parameterization, failure rate and repair time data. Wind farm-specific historical and expected cyclone conditions are constructed using a real-world data-calibrated Holland model. A turbine blade fragility curve is compiled based on previous studies. A two-parameter Weibull distribution component degradation model is adapted to directly adopt real-world failure rate data. We take the 54-GW planned offshore wind farms in South China as a case study given its substantial risk due to large capacity and frequent tropical cyclones. Results show that historical tropical cyclones could reduce annual wind farm availability by up to similar to 14 percentage points and induce a long-term declining trend. Excluding structural damage, maintenance interruptions alone under expected tropical cyclones with return periods exceeding the turbine design lifetime still lead to substantial availability losses. Our work provides a holistic framework for quantifying tropical cyclone risks, reveals their substantial and intensifying impacts, and brings attention to the considerable but often overlooked threat from maintenance interruptions.

Signal intersections are key nodes in urban road traffic networks, and real-time queue length information serves as a core performance indicator for formulating effective signal management schemes in modern adaptive traffic signal control systems, thereby enhancing traffic efficiency. In this study, a roadside Light Detection and Ranging (LiDAR) sensor is employed to acquire 3D point cloud data of vehicles in the road space, which acts as an important method for queue length detection. However, during queue-length detection, vehicles in different lanes are prone to occlusion because of the straight-line propagation of laser beams. This paper proposes a queue-length detection method based on variations in vehicle point cloud features to address the occlusion of queue-end vehicles during detection. This method first preprocesses LiDAR point cloud data (including region-of-interest extraction, ground-point filtering, point cloud clustering, object association, and lane recognition) to detect real-time queue lengths across multiple lanes. Subsequently, the occlusion problem is categorized into complete occulusion and partial occlusion, and corresponding processing is performed to correct the detection results. The performance of the proposed queue length detection method was validated through experiments that collected real-world data from three urban road intersections in Suzhou. The results indicate that this method's average accuracy can reach 99.3%. Furthermore, the effectiveness of the proposed occlusion handling method has been validated through experiments.

In recent years, the shipping industry has undergone significant development, accompanied by substantial consumption of fossil resources. This article studies the energy dispatch problem of the ship-integrated energy system (S-IES) to improve environmental awareness. It aims to reduce carbon emissions, avoid unnecessary resource misallocation, and rely on fewer communication resources during the execution of its navigational duty. An energy dispatch model is designed to increase economic and environmental benefits. Secondly, a distributed optimal prescribed-time algorithm based on the event-triggered mechanism (D-PET) is proposed to search for energy dispatch solutions. It ensures a quick and accurate response to the fluctuating load demands of S-IES. In addition, D-PET eliminates the disadvantage of limited communication interaction resources among energy participants in S-IES. Finally, to validate the effectiveness of the proposed energy dispatch strategy, a simulated navigational duty is conducted using a training vessel on a route from Queensland Port to Victoria Port via New South Wales Port.

Non-contact human-machine interaction technologies have demonstrated broad application potential in intelligent wearable devices. Existing capacitive proximity sensors face limitations, including restricted sensing distance, high inter-unit crosstalk, and unstable motion tracking. This paper proposes a non-contact capacitive sensor featuring a wave-shaped interdigitated electrode structure and dual shielding layers. The proposed wave-shaped design effectively enhances sensitivity compared with conventional rectangular interdigitated electrodes. By combining windowed shielding layers with cross-layer routing, inter-unit crosstalk and parasitic capacitance are effectively suppressed. A 5 × 5 capacitive sensor array, together with a high-speed data acquisition circuit, enables multi-channel real-time sampling at approximately 100 Hz. A two-stage algorithmic framework is also introduced to enable highly accurate handwritten-letter recognition and support a wearable non-contact handwriting interaction system. Experimental results demonstrate that the proposed sensor system exhibits high proximity sensitivity, producing a maximum capacitance decrease of 0.36 pF in response to the approach of a human hand, while achieving a sensing distance exceeding 35 mm. The sensor features a fast response time of 51 ms and strong immunity to interference. For in-air handwriting, the system recognizes 10 mixed-case letters with 98.38% accuracy and 26 uppercase letters with 97.14% accuracy. These results indicate significant application potential in non-contact control, human-machine interaction, and intelligent input systems.

In the context of escalating global carbon emissions and rising costs of traditional fossil fuels, the singular reliance on fossil fuel-based equipment for energy supply is no longer sustainable. Consequently, an integrated energy system incorporating both renewable energy sources and conventional fossil fuel equipment, known as the Ship-Integrated Energy System (S-IES), has emerged. As a novel ship energy architecture, the S-IES enables the efficient synergistic utilization of heterogeneous energy sources. To ensure the safe, reliable, and green operation of the S-IES, an energy management model is first established. This model facilitates the synergistic optimization of economic and environmental performance during voyages, namely, the minimization of operational costs and total carbon emissions. Furthermore, to quantify the carbon reduction effectiveness of the constructed energy management model and to align with the carbon emission reduction regulations for the shipping industry drafted by the International Maritime Organization (IMO), this paper establishes a ship carbon emission intensity grading mechanism based on the Energy Efficiency Operational Indicator (EEOI). This mechanism quantifies the carbon emission intensity throughout the entire voyage, contributing to reduced total ship carbon emissions while, to a certain extent, safeguarding the economic benefits of shipowner. Finally, the effectiveness of the proposed S-IES energy management model and the carbon emission intensity grading mechanism is validated using the fundamental data of a bulk carrier as a case study.

Offshore wind energy is a widely applied clean energy source and an important pathway to achieve port decarbonization. It ensures the sustainable development of the marine economy. This paper proposes a distributed economic dispatch strategy for port integrated energy systems considering offshore wind power to reduce port carbon emissions and promote sustainable development. Firstly, considering the distributed characteristics of heterogeneous energy supply equipment in ports, a Port Integrated Energy system is established. Secondly, the transmission losses exist in the power lines of offshore wind farms, an optimal schedule model of PIES considering offshore wind farms is developed. Then, based on distributed mixed-integer linear programming, a distributed algorithm is proposed. Finally, the effectiveness of the proposed method is verified through simulation.

In many seaports, tugboats not only assist vessels in berthing/unberthing and escort vessels through restricted channels, but also transport pilots on/off vessels. However, existing studies usually simplify tugboat operations and rarely consider pilot-transfer duties within the joint optimization of vessel scheduling and tugboat assignment. This omission may misestimate tugboat availability, interrupt service continuity among pilot transfer, escort, and berthing/unberthing assistance, and increase vessel waiting time, tugboat idling, and fuel consumption. To address this problem, this study develops a joint optimization framework for vessel scheduling and multi-task tugboat assignment. A bi-objective mixed-integer nonlinear programming model is formulated to minimize total vessel time in port and tugboat fuel consumption, and an adaptive dual-population memetic large neighborhood search hybrid algorithm (ADP-MLNS) is developed to solve the model efficiently. A case study based on real operational data from a Chinese seaport shows that the proposed approach reduces total vessel time in port by 22.73% and tugboat fuel consumption by 12.82% compared with the port's traditional scheduling rules. Computational experiments further demonstrate that ADP-MLNS achieves better solution quality and competitive computational efficiency than benchmark algorithms. This study provides an effective decision-support method for coordinated vessel traffic organization.

Although autonomous ships (AS) have made significant progress in collision-avoidance decision-making (CADM) in open waters, safe autonomous navigation in confluence waters remains challenging due to high traffic density and ambiguous multi-ship interactions. This paper proposes a novel paradigm for cooperative CADM by leveraging the understanding, reasoning, and interaction capabilities of Large Language Model (LLM) agents. Specifically, a multi-ship CADM model for confluence waters is established based on a Partially Observable Markov Decision Process (POMDP) to simulate navigational characteristics. Subsequently, a hierarchical coordination framework inspired by human decision-making practices, termed HLLM-NAV, is developed. A central LLM agent—analogous to Vessel Traffic Service (VTS) operators—aggregates situational information and determines passage sequencing, while ship-level LLM agents integrate the coordinator's guidance with local state observations to generate executable collision-avoidance maneuvers through Chain-of-Thought reasoning. Furthermore, a Retrieval-Augmented Generation (RAG) knowledge base is constructed to endow LLM agents with continuous learning capabilities. Simulation results demonstrate that HLLM-NAV achieves safe and efficient navigation in representative confluence waters. Extensive evaluations further indicate strong adaptability and flexibility, substantially improving CADM safety for AS. This work provides a practical methodological pathway for integrating LLM capabilities into maritime decision systems and enabling highly autonomous ships operation in confluence waterways.

Accurate and timely vessel traffic information is essential for safe port operations, yet it is often unavailable or delayed. To address this gap, we propose an intelligent traffic organisation service (ITOS) architecture based on e-navigation. ITOS consists of three layers: the data layer for maritime data collection and storage, the business layer for traffic organisation modelling and intelligent service generation and the user layer for vessel interaction and service delivery. This architecture standardises real-time, on-demand exchange of digital traffic information, integrating resources across different channels. A case study at Huanghua Port, Bohai Sea, demonstrated that ITOS effectively provided intelligent traffic services, reducing congestion in space and time. By improving channel utilisation and minimising vessel waiting, ITOS enhances navigational efficiency and port operations. These results highlight ITOS's potential to support vessel traffic services, assist port authorities and strengthen maritime trade competitiveness.

With increasing the awareness of environmentally friendly, the ship-integrated energy system (S-IES) combined with power and heating networks has become an upcoming trend in the shipbuilding industry. It decreases the consumption of fuel-based resources by integrating renewable energy resources to replace traditional fossil-based generators partly. Meanwhile, with many intelligent devices, the S-IES performs partitioned and autonomous features because of the unique operational structure. Therefore, the normal operation of S-IES is highly vulnerable, which will be bothered by the infeasibility-driven attack for shipboard devices and influenced by the uncertainties of renewable energy resources. To this end, the energy management problem (EMP) under cyber attacks for S-IES is focused on ensuring secure and green sailing in this article. To balance the number of nods and decrease the communication resources among energy devices, a partitioning algorithm is designed to construct the energy router (ER)-based energy hub (EH) for S-IES. Meanwhile, considering the tradeoff relationship between the economy and carbon emission, a multiobjective energy management model with feasible sailing constraints is designed. Then, to ensure the information exchange among EHs using ERs, a distributed resilient energy management strategy with a detection mechanism is proposed to acquire the optimal solutions under cyber attack. Finally, the theoretical analysis verifies the effectiveness of the designed distributed resilient strategy. Simulation results of different scenarios prove the great performance on ensuring secure and green sailing for ships by the constructed energy management model and its distributed resilient strategy.

Unmanned boats will encounter many static and dynamic obstacles during navigation, and only real-time obstacle sensing can ensure safe navigation and long endurance of unmanned boats. In this paper, LGCGNet is proposed to perform real-time water surface semantic segmentation on the images captured by the on-board camera. In order to ensure that the model adapted to obstacles with extremely variable scales, a local-global module is proposed in this paper. The local-global module consisted of residual dense dilated module and context-enhanced separable self-attention. Residual dense dilated module enabled the enhancement of local detail information and context-enhanced separable self-attention enabled model receptive field expansion. In addition, the sub-pixel downsampling module is used to avoid the loss of feature information to improve segmentation accuracy. Experiments on the MaSTr1325 dataset showed that LGCGNet apprpached the segmentation accuracy of state-of-the-art semantic segmentation models with only 689,000 parameters and 9.068G floating point operations per second, with an mIoU of 84.14

To reduce carbon emissions in the shipping industry, the energy management problem of the ship-integrated energy system (S-IES) is analyzed in this paper. Firstly, a marine environmental risk field model is constructed to quantify the degree of hazard when designing the sailing route. Meanwhile, an energy management model that considers both economic and environmental benefits is developed to enhance the penetration rate of renewable resources. Subsequently, a distributed energy management algorithm based on finite-time consensus theory is proposed to ensure a rapid and accurate response to load demand. Moreover, a mathematical analysis is provided to demonstrate the algorithm’s effectiveness. Finally, the sea area between Singapore Port (Singapore) and Penang Port (Malaysia) is chosen as the simulation environment. The experimental results demonstrate the effectiveness of energy management for the S-IES.

With the rapid development of Internet of Things (IoT) technology, unmanned surface vehicles (USVs), as critical nodes in distributed maritime sensing networks, have seen their autonomous collision avoidance capabilities become central to enabling coordination among multiple devices and real-time decision-making. Based on this, this article presents an intelligent strategy for collision and obstacle avoidance in USVs during multitarget encounter scenarios. A collision risk model based on the International Regulations for Preventing Collisions at Sea (COLREGs) and common practices of sailors is constructed, with it being used as a constraint for the differential evolution (DE) algorithm. The objective function is decomposed geometrically and in terms of states, transforming the evaluation of the entire path into evaluating individual path points. In this way, high-quality path points are fully utilized, and a fitness function is built based on each individual path point. The population initialization operation of the DE algorithm is improved through a chaotic multipopulation parallel optimization strategy, with a chaotic matrix being introduced to enhance search traversal. Additionally, a parameter randomization strategy is introduced in mutation and crossover operations to avoid local optima, and each subpopulation is optimized in parallel to obtain the best collision avoidance route. Finally, the simulation experiments results demonstrate that the improved DE algorithm demonstrates superior performance in both collision avoidance efficiency and path optimization, confirming the effectiveness of the approach in complex multitarget encounter scenarios.