Understanding human motion is fundamental to the development of intelligent systems capable of seamless interaction with people. Trajectory prediction is a critical component in domains, such as intelligent transportation, surveillance, and human–robot collaboration. However, accurately forecasting human movement remains a significant challenge due to its inherently uncertain and multimodal nature. In this work, we propose a deep neural network that models agent dynamics and predicts future trajectories by representing them as a probabilistic multimodal distribution. To effectively capture the stochasticity of human behavior, our method employs a score-based diffusion model that learns to generate realistic trajectory samples by denoising latent representations. In addition, we introduce a novel social attention mechanism designed to model complex interagent interactions, further improving predictive performance. We validate our approach in both pedestrian and marine vessel trajectory datasets, demonstrating its superior ability to capture social dynamics and forecast diverse plausible future outcomes. Extensive experiments and ablation studies confirm the robustness, generalizability, and accuracy of our framework in varied real-world environments.

Some advanced methods have leveraged the zero-shot recognition capability of the contrastive language–image pretraining (CLIP) model and adapted it to weakly supervised semantic segmentation (WSSS), achieving promising performance. However, they primarily use CLIP as an auxiliary feature extractor, leaving the fundamental limitations of class activation mapping unresolved, particularly in preserving fine-grained object boundaries and achieving precise pixelwise localization under sparse supervision. To address these challenges, this article proposes a novel end-to-end WSSS framework WTCLIP, which aims to fully exploit the potential of CLIP for weakly supervised segmentation tasks. Different from traditional methods that use CLIP only as a static feature extractor, we innovatively introduce a learnable wavelet transform decoder to enhance the information extraction capability and significantly improve the model's perception of object boundaries. We dynamically adjust the weight distribution ratio of the CLIP feature layer, capture multiscale edge information, and make full use of the time–frequency localization characteristics of the wavelet transform to significantly improve the quality of pseudolabels and achieve more accurate semantic segmentation. Experimental results show that our method significantly improves the performance of the WSSS task on two public benchmark datasets, notably by4.0%over the state-of-the-art methods, especially in capturing weakly annotated object boundary details.

Battery systems are increasingly utilized in oceangoing ships, with a growing number of fully electric or hybrid vessels relying on battery power for propulsion. Ensuring the safety of these ships necessitates continuous monitoring of the available energy storage within the batteries. Classification societies typically mandate that the state of health (SOH) be verified through independent tests. Effective monitoring and maintenance of maritime battery systems are crucial for the sustainability and safety of marine operations. This study proposes a novel approach to predict the SOH using dual-perspective 1D fully convolutional neural networks (1D-FCNs) and snapshot data processing techniques. This paper aims to enhance predictive accuracy and model robustness by integrating snapshot curves extracted from operational data into the proposed dual-perspective FCN model. The snapshot method is notable for its ability to capture the instantaneous state of the battery through various operational parameters. Unlike traditional cumulative methods, which require extensive historical operational data, the proposed approach only necessitates specific operational data with the required features. This reduces processing time, data transfer demands, and infrastructure costs, which are particularly critical for maritime applications.

Accurately predicting the remaining useful life (RUL) of lithium-ion batteries plays a crucial role in the sustainable development and efficient operation of fields such as new energy vehicles, numerous electronic products, and energy storage power stations. This study presents a dual-exponential model integrated with an Extended Kalman Filter (EKF) for RUL prediction, enhanced by Bayesian optimization to automatically adjust the model parameters for improved accuracy. A dynamic weight loss function is employed to adapt to battery characteristics at various degradation stages, enabling optimal model parameterization. The model is evaluated using four datasets under different starting prediction cycles (300, 400, and 500 cycles). Experimental results demonstrate that the model effectively tracks capacity degradation and predicts RUL with high accuracy, fitting well with actual degradation curves and showing strong generalization ability across various datasets. Based on the conducted experiment, the proposed model demonstrates improved accuracy in tracking capacity degradation and predicting RUL.

Current cross-domain few-shot semantic segmentation (CD-FSS) faces multiple challenges, including inconsistent feature mapping among domains and insufficient utilization of low-level information and background information from the source domain. To address these issues, this article proposes a novel cascade feature enhancement and contrastive learning framework to improve the generalization capability of CD-FSS. Within this framework, we first introduce a cascade feature enhancement module to construct distinctive feature representations, enhancing the model’s transferability across domains. By effectively integrating multilevel feature information from support images, this module strengthens the representation capability of query images. Second, we employ contrastive learning to form positive and negative sample pairs for the foreground and background, capturing rich correlations between them. Finally, the iterative prototype enhancement module we propose gradually refines the correspondence between the support image and the query image through iteration, making full use of the embedded supervisory information in the limited support samples. Experimental results demonstrate that the proposed method outperforms existing approaches on multiple benchmark datasets, achieving up to a 9.7% improvement over state-of-the-art methods.

The safety of navigation is critical in areas with heavy and complex traffic. Accurate prediction of the future trajectory of ships is crucial, especially in encounter situations where conventional navigational devices are prone to high uncertainty and risk due to their limitations. Unfortunately, kinematic models and data-driven methods suffer from numerous issues, such as poor performance, intricate architecture, and reduced interpretability. In response, this article introduces a novel approach, driven by automatic identification system (AIS) data, to enhance the efficacy of ship encounter trajectory prediction. It revolves around encounter classification, wherein the invaluable insights of seamanship play a pivotal role in identifying and categorizing legitimate encounters into three distinct types. This categorization forms the foundation for developing a probability-based classification model, in conjunction with a hybrid predictor that amalgamates a kinematics-based model and a neural network-based model. Historical AIS data collected in Oslofjord, Norway, are utilized in this study, and experiments have been conducted to assess the performance of the proposed method through real cases. The results substantiate the promise of the classification model and underscore the exceptional predictive capabilities of the hybrid approach. This superiority is attributed to the synergistic effects arising from the integration of its constituent models.

Navigation safety is of paramount importance in areas with heavy and complex maritime traffic. Any ship navigating such a scenario should be able to foresee the future positions of other ships and adjust its path accordingly to avoid collisions. However, predicting future trajectories is a very challenging problem due to many possible future trajectories from the inherent uncertainty and the complex interaction dynamics between different ships. In this article, we propose a deep generative model based on the conditional variational autoencoder framework to learn marine vessel movement and predict future trajectories. The model is able to produce a multimodal probability distribution over future trajectories and model the complex interactions between vessels. Experiments are performed in two-vessel encounter scenarios from real-world automatic identification system data. The proposed model outperforms the baseline methods, including both kinematics-based and data-driven methods. The trajectories predicted by the proposed model are also analyzed to demonstrate the effectiveness of the model.

To decrease the risk of collisions and ensure safe navigation, an Automatic Identification System (AIS) was developed to broadcast real-time ship states information, such as ship position and sailing speed. Due to the real-time characteristic of AIS data, it, hence, has been widely used to construct a data-driven model for ship trajectory prediction, which can provide onboard decision support for navigators. However, data provided by AIS is sometimes partially missing or simply wrong. Poor quality of AIS data can degrade the performance of data-driven models and lead to large uncertainty in the predicted ship positions. Therefore, this paper proposes a hybrid model to reduce uncertainty by integrating the multi-output Gaussian process (MOGP) model predictions and historical trajectory information. Historical trajectories provide useful prior knowledge to calibrate the performance of MOGP model predictions. This model is built through three steps: 1) extracting historical ship trajectory information from the route that a ship is following; 2) predicting ship positions with a data-driven predictor built by an MOGP; and 3) obtaining a trajectory by knowledge fusion of historical information and predicted results. The experiment shows the proposed hybrid model outperforms other data-driven models and has small errors and uncertainty quantification of the ship position prediction. Especially for the 6-minute position prediction, its RMSE is over 100 meters smaller than other methods.

Estimating oceanic wave fields from sparse observations has been a long-standing challenge in oceanography and an important environmental metric desired for maritime operations. The requirement for frequent real-time updates of the wave field within the local area poses difficulties for data assimilation approaches, as they can be computationally complex and rely on external atmospheric forcing. The relationship between the wave field and local sparse observations is embedded in reanalysis or hindcast data. We propose a data-driven deep-learning model capable of estimating the local wave field using sparsely distributed floating wave buoys. This novel model simultaneously produces wave height, period, and direction, along with their respective uncertainties. In a year-long test period within a local fjord region characterized by complex wave patterns influenced by intricate geography, the proposed model demonstrates remarkable accuracy and efficiency in estimating wave fields. This study demonstrates the promising potential of data-driven deep-learning models as an alternative to rapidly estimating the wave field.

The visual attention of navigators is imperative to understand the logic of navigation as well as the surveillance of navigators' status and operation. Current studies are implemented with the help of wearable eye-tracker glasses; yet, the high expenditure demanded by such equipment and service and its limitations on usability have impeded related research from further development. In this letter, the authors propose a framework, which is the first attempt in the maritime domain, to provide a camera-based deep-learning (CaBDeeL) visual attention recognition solution that outperforms the intrusive eye tracker regarding its shortcomings. A wide-angle camera is configured in front of the navigator in the advanced ship-bridge simulator in a way that visual attention reflected by their facial and head movements is captured in the front view. A pair of eye-tracker glasses is used to classify the captured visual attention images, which then form the primary database. During the process of classifying camera-captured images, a convolutional neural network (CNN) is built as an automatic classifier. The CNN is applied to two scenarios, and it shows an overall 95 % accuracy.

Understanding current environmental conditions is essential for autonomous ships, among which real-time estimation of sea conditions is a key aspect. Considering the ship as a large wave buoy, the sea state can be estimated from motion responses without extra sensors installed. This task is challenging since the relationship between the wave and the ship motion is hard to model. Existing methods include a wave buoy analogy (WBA) method, which assumes linearity between wave and ship motion, and a machine learning (ML) approach. Since the data collected from a vessel in the real world are typically limited to a small range of sea states, the ML method might fail when the encountered sea state is not in the training dataset. This article proposes a hybrid approach that combines the above two methods. The ML method is compensated by the WBA method based on the uncertainty of estimation results, and thus, the failure can be avoided. Real-world historical data from the Research Vessel Gunnerus are applied to validate the approach. Results indicate that the hybrid approach improves the estimation accuracy.

Sea state estimation is beneficial for marine systems to enhance on-board decision-making and improve work efficiency. In the era of ship intelligence, artificial intelligence has greatly promoted the technology of sensing environment, such as by using the deep learning. However, it is difficult to collect enough motion data from a marine system to train a deep learning model. In addition, the model for sea state estimation is trained using the data from a specific marine system; applying the model directly to another marine system may result in performance degradation. In this paper, a supervised transfer learning based framework for sea state estimation (STLSSE) is proposed. The STLSSE focuses on knowledge transfer when the collected data for the source marine system is sufficient but the collected data of the target marine system is scarce. In STLSSE, a data pairing algorithm is proposed to determine the relationship of the source and the target marine system. Based on these paired data, a Siamese convolutional neural network, including a new proposed residual fully convolutional network and two novel attention modules, is designed for the semantic alignment. Moreover, the conventional contrastive loss is improved to characterize the distributions when there are only few samples in the target marine system. The extensive comparisons between STLSSE and state-of-the-art transfer learning approaches show its superior performance. The comparisons with state-of-the-art attention modules has verified the competitiveness of the proposed attention modules. The key parameters and each component of STLSSE are emphasized in the ablation and sensitivity studies.

The advent of the COVID-19 pandemic disrupted global commercial activities and the tourism industry heavily. Impacts on maritime transportation were huge, as seaborne trade represents over 80% of global merchandise trade. Investigating how COVID-19 has affected ship behaviours is significant for economic condition evaluation, port management. This paper develops an analysis method to mine knowledge from raw Automation Identification System (AIS) data. First, berths are identified by improved density-based spatial clustering of applications with noise by Pythagoras distance (PD-DBSCAN). Data features, such as ship deadweight, arrival time, dwelling time, ship types, etc., can then be extracted using information matching and statistical analysis. Next, the dynamic time warping method is employed to analyse abnormal ship behaviour patterns and quantify the impacts of COVID-19. After that, a significance test is employed to determine an impact threshold through year-on-year analysis on ship flow, daily throughout and berthing time of quays. Finally, statistical analysis is used for the short-term impact analysis. This research examines a case study based on four-year AIS data in the Oslo port area. The results show that the proposed method can identify abnormal patterns caused by COVID-19 and estimate its impacts. Passenger ships are influenced heavily compared with cargo ships. The variation of passenger ships’ flow is over 90% during 2020, larger than the average variation before 2020. The discovered knowledge could be used for future decision-making and preplanning in the next health crisis.

Situation awareness is of great importance for autonomous ships. One key aspect is to estimate the sea state in a real-time manner. Considering the ship as a large wave buoy, the sea state can be estimated from motion responses without extra sensors installed. However, it is difficult to associate waves with ship motion through an explicit model since the hydrodynamic effect is hard to model. In this paper, a data-driven model is developed to estimate the sea state based on ship motion data. The ship motion response is analyzed through statistical, temporal, spectral, and wavelet analysis. Features from multi-domain are constructed and an ensemble machine learning model is established. Real-world data is collected from a research vessel operating on the west coast of Norway. Through the validation with the real-world data, the model shows promising performance in terms of significant wave height and peak period.