In dynamic and cognitively demanding human–machine environments, achieving reliable system performance across diverse domains is critical for ensuring ergonomic interaction and cognitive efficiency. This article introducesChannel-filtered Learning with Ensemble Attention and Robustness(CLEAR), a domain generalization framework that simultaneously suppresses domain-specific feature interference and enhances domain-invariant representation learning. CLEAR integrates ensemble multiscale spatial attention with a progressive channel filtering mechanism to selectively eliminate features most indicative of domain identity. Furthermore, a momentum-based inference consistency loss is introduced to promote semantic stability by aligning class prototypes across domains over time. These components operate in concert to enable consistent task-relevant feature extraction and improved generalization under unseen distributional conditions. Extensive evaluations on multiple public benchmarks demonstrate that CLEAR achieves state-of-the-art performance in domain generalization while enhancing robustness and cognitive reliability in real-world human–machine systems. The framework contributes to the development of perceptually ergonomic AI models capable of maintaining accuracy across varying operational contexts.

The proliferation of Internet users has accelerated the spread of fake news on social media, necessitating effective fake news detection systems. However, existing methods often focus solely on the features of claims without considering their uncertainty, limiting their reliability and generalizability. Inspired by Bayesian neural networks, this paper proposes a Bayesian Transformer-based Fake news Detection system with Evidence awareness (BTFDE). To validate the effectiveness of BTFDE, we conducted experiments on several datasets. The results demonstrate that BTFDE outperforms several baseline methods, improving the reliability of fake news detection by quantifying uncertainties and incorporating evidence awareness. This approach improves the generalizability of the model and provides a more rational basis for fake news detection.

This work presents a novel Multi-Scale Feature Aggregation Filtering and Consistency Loss (MFAFC) to improve the extraction of domain-invariant features, crucial for enhancing human-machine system adaptability in dynamic environments. Unlike existing techniques that primarily focus on domain-invariant features while neglecting the influence of domain-specific features, this approach considers both aspects. By capturing information across multiple scales, it suppresses domain-specific features, improving domain-invariant feature extraction quality. A momentum-based inference consistency loss function is also introduced, using category center consistency to boost model robustness. Combining multi-scale extraction and momentum-based loss, the method effectively handles domain shift. Experiments on various public datasets show excellent performance in domain generalization, reducing the impact of domain-specific features and improving task ergonomics and cognitive performance in real-world systems.

In modern aviation and logistics, the professional competence and operational skills of unmanned aerial vehicle (UAV) operators are critical to mission success and overall efficiency. Accordingly, effective UAV training, which ensures high levels of involvement, is paramount in enhancing operators’ learning and operational performance. To address the lack of theoretical research on electroencephalography (EEG)-based involvement prediction for UAV training, this study proposes EEG-IvNet, a deep learning framework designed to predict involvement from EEG signals. The model integrates convolutional neural networks (CNNs), long short-term memory (LSTM) networks, and an attention mechanism to classify involvement states. EEG data were collected from participants performing UAV monitoring tasks in controlled experiments and used as input to the model. The results demonstrate that EEG-IvNet outperformed the benchmark CNN+LSTM model in classifying involvement states. Furthermore, this study offers insights into the neural basis of involvement, providing strong theoretical and algorithmic support for incorporating VR technology in UAV training. These findings not only suggest potential improvements in training outcomes but also highlight broader applicability in intelligent aviation, virtual education, and other related domains.

In future battlefields, missiles are expected to become highly precise and efficient strike weapons, with missile intelligence emerging as a critical development trend. To address the problem of optimizing 3-D missile interception guidance laws, this article introduces the deep Q-network (DQN) algorithm on the foundation of proportional navigation guidance (PNG) and proposes an adaptive proportional guidance algorithm based on deep reinforcement learning (DRL). The proposed algorithm uses air combat situational information as the state space and incorporates parameters such as the missile-target relative distance and line-of-sight (LOS) angle into the reward function design. The optimal proportional navigation coefficient$K^{*}$for low-overload maneuvering targets is determined through network search, and the longitudinal and lateral control commands of the missile are decoupled by designing the proportional coefficient increment$\Delta K$, constructing a discretized action space. Simulation results show that, compared to the PNG with a constant$K^{*}$, the proposed method significantly improves the hit probability of high-overload maneuvering targets while maintaining the hit rate for low-overload maneuvering targets. As an exploration of future intelligent combat scenarios, this guidance law design method holds both theoretical significance and practical application value.

For many image fusion problems, the most used technique is selecting features with rich information. The robust image fusion method based on block compressive sensing principle is studied here. Compressive sensing is known to provide an effective method with high accuracy. The framework of the proposed method is given in various perspectives: block sparse representations, restoration algorithms, feature extraction, online dictionary learning, and fusion rules. In terms of restoration of fused images, the split Bregman iteration is adopted. The proposed method can acquire well fusion image from source images and remove some degradations simultaneously, such as noises and blurring effect. In addition, both ‘maximum selection’ and ‘weighted mean’ are investigated as fusion rules, which can preserve more information. Generally, the proposed method can achieve better fusion result from the source images. The experiments with or without noise source images both illustrate that the proposed method has relatively comparative fusion results.

Conventional semi‐automatic or interactive methods, which require a small amount of user inputs for region segmentation of objects, have obtained the best segmentation results. A new semi‐automatic segmentation technique by using coloured scale‐invariant feature transform (CSIFT) to extract seed pixels in Graph Cuts is introduced here. First, CSIFT is used to extract feature points of objects in the image. Then, a voting process is used to extract the matched points as object seeds. The detailed technique via s–t Graph Cuts has been presented, and a new segmentation energy cost function with two colour‐invariant descriptors has been proposed: colour‐name descriptor and colour‐shade descriptor. The colour‐name descriptor introduces high‐level considerations resembling top‐down intervention, and the colour‐shade descriptor allows us to include physical consideration derived from the image formation model capturing gradual colour surface variations and provides congruencies in the presence of shadows and highlights in the segmentation. The experimental results prove that the proposed method provides high‐quality segmentations with object details.