Linyu Wang

Accurate vessel trajectory prediction is critical for maritime safety, traffic management, and autonomous navigation in complex waters. However, vessel movements are highly nonlinear, dynamically variable, and characterized by long-range spatiotemporal dependencies, which limit the performance of existing methods in both short- and long-term prediction. To address these challenges, this paper proposes BMAG, a vessel trajectory prediction framework that integrates Bidirectional Mamba2, Attention, and GRU. Built on AIS data, BMAG employs an end-to-end encoder-decoder architecture with an input embedding layer, a bidirectional Mamba2 encoder, and an autoregressive GRU decoder with multi-head cross attention. The encoder captures bidirectional spatiotemporal interactions from historical trajectories and fuses contextual information through a learnable weighting mechanism, while the decoder selectively exploits relevant historical states to improve prediction accuracy and stability. Experiments on 2024 AIS datasets from Tampa Bay and the Port of New York show that BMAG consistently outperforms advanced baselines, reducing ADE and FDE by up to 23.5% and 14.6% for 20-min prediction, and by 31.5% and 28.7% for 60-min prediction, respectively. These results highlight the effectiveness and robustness of BMAG for multi-horizon vessel trajectory prediction in complex maritime environments.

Spatiotemporal trajectory classification is essential for intelligent perception systems but faces challenges including weak separability of dynamic features, representation collapse under limited samples, and heterogeneous conflicts in multimodal data. To address these issues, we propose K-M LLM-pro, a physics-guided cross-modal adaptation framework that integrates statistical mechanics with large language models (LLMs) to improve trajectory understanding. Our approach incorporates: (1) physics-informed prompt engineering based on Kramers-Moyal coefficients, embedding physical constraints via reproducing kernel Hilbert space projection; (2) a dynamic patching optimization mechanism combining variance maximization and Lyapunov stability criteria for unified modeling of heterogeneous trajectories; and (3) dual spatiotemporal adapters with a parameter-efficient expansion strategy, injecting domain knowledge while optimizing only 3.8% of new parameters. Experimental results on public datasets such as Geolife and AIS show that K-M LLM-pro outperforms state-of-the-art models in classification accuracy, demonstrating strong performance even in few-shot scenarios with only 1% of training data. To our knowledge, this is the first work to integrate K-M coefficients as interpretable statistical priors into LLMs, offering a lightweight and effective solution for modeling complex spatiotemporal dynamics.

Accurate short-term flight trajectory prediction plays an important role in Air Traffic Control, which can ensure the safety of air transportation. Two main issues limit the prediction performance of existing works. The first issue is non-discriminative feature representation for flight trajectory attributes. The second issue is that real-world flight trajectories involve complex temporal dependencies, yet most existing methods extract features from a single time scale, failing to capture intricate temporal patterns. In this paper, we propose a framework (M2FlightNet) to address the above issues. M2FlightNet employs a differential representation for longitude and latitude to provide discriminative features and reflect the underlying temporal variations in trajectories. A mixture of experts trend extractor is adopted to capture trend features in vertical flight state series. To fully explore the diverse temporal patterns in horizontal flight state series, a multi-scale patch network is delicately designed to serve as the backbone. The multi-scale patch network exploits stacked patch mixer blocks to capture inter- and intra-patch dependencies under different time scales, and further integrates multi-scale temporal features across different scales and variables. After that, M2FlightNet ensembles multiple predictors to make direct multi-step prediction and reconstruct the trajectories. Extensive experiments on six real-world datasets demonstrate that M2FlightNet outperforms the state-of-the-art methods across various settings.

The intelligent identity authentication based on specific emitter identification (SEI) shows great promise in the field of military and civilian. In real-world scenarios, the distribution of labeled and unlabeled samples is highly imbalanced, only a few labeled samples are available. SEI methods based on consistency regularization (CR) have been proven effective in addressing such issue. However, most existing methods suffer from limited portability, in other words, they are only applicable to specific emitters in specific scenarios. In this letter, a robust framework based on revisiting data augmentation strategies and phased training (RDASPT) is proposed. Two data augmentation called hybrid augmentation (HA) and sampling augmentation (SA) are designed to address the issues of scarcity of labeled samples, poor availability of augmented samples and inconsistency of the predicted class distribution. The design of phased training is conducive to mining deep semantic features embedded in labeled and unlabeled samples. When only 5% of the labeled samples are available, the recognition accuracy reaches over 97% on two real-world datasets, which verify the advantages of RDASPT for data-scarce and portable SEI tasks.

The unmanned aerial vehicle (UAV) has drawn attention from the military and researchers worldwide, which has advantages such as robust survivability and execution ability. Mobility models are usually used to describe the movement of nodes in drone networks. Different mobility models have been proposed for different application scenarios; currently, there is no unified mobility model that can be adapted to all scenarios. The mobility of nodes is an essential characteristic of mobile ad hoc networks (MANETs), and the motion state of nodes significantly impacts the network’s performance. Currently, most related studies focus on the establishment of mathematical models that describe the motion and connectivity characteristics of the mobility models with limited universality. In this study, we use a backpropagation neural network (BPNN) to explore the relationship between the motion characteristics of mobile nodes and the performance of routing protocols. The neural network is trained by extracting five indicators that describe the relationship between nodes and the global features of nodes. Our model shows good performance and accuracy of classification on new datasets with different motion features, verifying the correctness of the proposed idea, which can help the selection of mobility models and routing protocols in different application scenarios having the ability to avoid repeated experiments to obtain relevant network performance. This will help in the selection of mobility models for drone networks and the setting and optimization of routing protocols in future practical application scenarios.

Background: Studies have proven that the same family of non-coding RNAs (ncRNAs) have similar functions, so predicting the ncRNAs family is helpful to the research of ncRNAs functions. The existing calculation methods mainly fall into two categories: the first type is to predict ncRNAs family by learning the features of sequence or secondary structure, and the other type is to predict ncRNAs family by the alignment among homologs sequences. In the first type, some methods predict ncRNAs family by learning predicted secondary structure features. The inaccuracy of predicted secondary structure may cause the low accuracy of those methods. Different from that, ncRFP directly learning the features of ncRNA sequences to predict ncRNAs family. Although ncRFP simplifies the prediction process and improves the performance, there is room for improvement in ncRFP performance due to the incomplete features of its input data. In the secondary type, the homologous sequence alignment method can achieve the highest performance at present. However, due to the need for consensus secondary structure annotation of ncRNA sequences, and the helplessness for modeling pseudoknots, the use of the method is limited. Results: In this paper, a novel method “ncDLRES”, which according to learning the sequence features, is proposed to predict the family of ncRNAs based on Dynamic LSTM (Long Short-term Memory) and ResNet (Residual Neural Network). Conclusions: ncDLRES extracts the features of ncRNA sequences based on Dynamic LSTM and then classifies them by ResNet. Compared with the homologous sequence alignment method, ncDLRES reduces the data requirement and expands the application scope. By comparing with the first type of methods, the performance of ncDLRES is greatly improved.

Evidence has accumulated enough to prove non-coding RNAs (ncRNAs) play important roles in cellular biological processes and disease pathogenesis. High throughput techniques have produced a large number of ncRNAs whose function remains unknown. Since the accurate identification of ncRNAs family is helpful to the research of their function, it is of necessity and urgency to predict the family of each ncRNAs. Although several traditional excellent methods are applicable to predict the family of ncRNAs, their complex procedures or inaccurate performance remain major problems confronting us. The main idea of those methods is first to predict the secondary structure, and then identify ncRNAs family according to properties of the secondary structure. Unfortunately, the multi-step error superposition, especially the imperfection of RNA secondary structure prediction tools, maybe the cause of low accuracy. In this paper, a novel end-to-end method `ncRFP' was proposed to complete the prediction task based on Deep Learning. Instead of predicting the secondary structure, ncRFP predicts the ncRNAs family by automatically extracting features from ncRNAs sequences. Compared with other methods, ncRFP not only simplifies the process but also improves accuracy. The source code of ncRFP can be available at https://github.com/linyuwangPHD/ncRFP.