To overcome the problem of low system resource utilization caused by the lack of exploration of environmental changes in Industrial Internet of Things (IIoT) service orchestration, while ensuring Quality of Service (QoS), we propose an Adaptive Orchestration Algorithm of Service Function Chain (AOASFC) Based on Deep Reinforcement Learning (DRL). Our approach in paper integrates joint deployment and routing information to manage system resources, thereby optimizing orchestration strategies. Furthermore, to enhance the capability of exploring environmental changes, we design a curiosity-driven module that evaluates the environmental changes before and after the DRL agent’s decision-making process, generating intrinsic rewards to guide a more comprehensive exploration process. Our approach effectively mitigates the high bias issue caused by updating value function, because we integrate Proximal Policy Optimization (PPO) with Generalized Advantage Estimation (GAE) and perform weighted averaging on multi-step estimates, optimizing temporal difference learning. In performance comparisons, we have compared DeepCoordblue(a centralized DRL orchestrator) and BSP(a greedy heuristic baseline) algorithm, AOASFC demonstrates superior performance in different traffic arrival patterns of SFC deployment scenarios: it not only improves system throughput by 15.35% and 11.64% respectively, but also keeps end-to-end latency below 50ms while significantly enhancing resource utilization.

In Internet of Things (IoT)-enabled marine sensor networks, underwater object detection faces challenges due to resource constraints, such as small object identification and scale variations. These challenges result in sample imbalance and ambiguities in label assignments. While frameworks like the YOLO series are efficient, their underwater performance is often inadequate due to these issues. This paper introduces a training enhancement strategy tailored for underwater object detection (UWDET) in IoT settings, aimed at reducing inference resource consumption while maintaining high detection accuracy. Importantly, our approach preserves existing network architectures and does not extend inference time. The methodology comprises three main elements: Gaussian Overlap Loss (GOL), which interprets bounding boxes through two-dimensional Gaussian distributions, thereby enhancing localization and addressing scale imbalance for small objects in resource-limited environments. Dynamic Task Joint Assignment (DTJA) modifies positive sample assignments based on classification confidence and regression quality, thereby minimizing false positive rates during training. Normative Focal Loss (NFL) employs a normalized joint assignment metric as continuous labels to effectively address sample imbalance. Experimental evaluations on underwater detection benchmarks reveal that our approach markedly enhances precision and recall, stabilizes gradient signals, and improves training efficiency. We also report training-side GPU memory/time/energy and edge-side memory and latency, confirming unchanged inference cost and reduced training resource usage. Our training enhancement strategy applies meticulously designed lightweight generic object detection models to the underwater domain. Without requiring complex modifications to the network architecture, it enables rapid training and facilitates the deployment and inference of high-precision models within resource-constrained IoT underwater devices.

Medical Mixed Reality (MR) has made significant progress in virtual surgery simulation and tumor teaching. This paper proposes a framework for pulmonary nodule attribute editing based on image feature consistency, achieving spatial alignment of multi-stage case data. To address the limitations of traditional time-image reconstruction, we design an adversarial siamese model architecture capable of synthesizing missing nodule images, completing temporal data, and fine-grained modeling of nodule growth. To tackle challenges such as deformation, background inconsistency, and attribute uncertainty in generated samples, we introduce a Denoising Diffusion Implicit Model (DDIM) and construct an attribute vector space for pathological feature editing. Additionally, we propose a separable image reconstruction strategy to enhance local feature stability. Extensive validation on the lung-specific LIDC-IDRI dataset demonstrates superior performance with SSIM of 97.5${\%}$ and LPIPS of 0.036. To further verify generalization capability, cross-organ testing on the liver-focused LiTS dataset achieves competitive results with SSIM of 85.0${\%}$ and LPIPS of 0.128. These outcomes provide strong technical support for high-fidelity virtual surgery and intelligent tumor teaching platforms.

The CNN-Transformer structure is widely applied to underwater image enhancement (UIE) tasks. However, previous studies have typically used structures similar to those in other image restoration scenarios, without specifically designing a unified structure that fully integrates the characteristics of convolution and Transformers for real-world underwater scenarios. Moreover, the inconsistency of color channel and spatial region attenuation of underwater images has not been given sufficient attention. To this end, this paper proposes a new UIE network, UMCTN, based on multi-kernel convolution Transformer. A multi-kernel convolution residual self-attention block (MCRA) was constructed. By designing a multi-kernel convolutional residual structure, it addresses the issue of critical feature information loss when convolution is applied to small-sized image patches in Transformers, which are widely adopted to reduce computational costs. It elegantly combines the characteristics of convolution and Transformers, endowing the network with a strong capability to capture both local and global dependencies. In addition, a feature fusion compensation module (FFCM) is proposed to supplement richer global perceptual features for MCRA, and It can effectively remove and restore spatial and color channels with more severe attenuation. Tests on the UIEB, UFO-120, and LSUI datasets show that UMCTN achieves better quantitative evaluation and visual performance compared to state-of-the-art (SOTA) schemes, with a maximum peak signal-to-noise ratio (PSNR) improvement of 1.93 dB. Detailed ablation studies validate the effectiveness of each component.

In industrial Internet of Things, the combination of specific emitter identification (SEI) and key authenticated technologies can effectively resist spoofing attacks and improve system security. However, most existing SEI approaches extract features based on real valued operations and only work in static scenario. This motivated us to develop a novel SEI method tasked with: exploiting the high potential model for SEI based on the inphase/quadrature (I/Q) signal that is represented by complex number, and realizing rapid reconstruction of the model in the face of dynamic scenarios. To this end, in this article, we introduce a two-stage cross-domain identification model. First, a complex-valued separate residual network (CVSRN) with novel separate residual modules is proposed as the pretrained model. The CVSRN can automatically extract effective inherent features directly from raw signals in an end-to-end manner, which favors complex-valued signals that are found in two distinct signal paths. Second, three transfer strategies are proposed to achieve rapid construction of the target SEI model. They leverage the knowledge learned from the pretrained CVSRN to facilitate the recognition of a new but similar emitters. We benchmark our proposed approach against four state-of-the-art SEI methods on real-world data and exhibit that it is not only competitive but also able to cope with complex dynamic scenarios.

Remaining useful life (RUL) prediction of aero-engines is one of the important issues in research related to engine health management. Although deep learning has made great progress in fault diagnosis research, successful training of deep learning models is very time-consuming and difficult to meet the real-time requirements of online RUL prediction applications. Broad learning systems (BLS) provide an alternative to deep learning networks with low computational resource requirements, fast training time, and incremental scalability. Based on the typical BLS, we propose a new lightweight multiscale BLS (MSBLS). Considering that RUL is influenced by the working condition factor, the discrete wavelet transform is used to generate multiresolution components, and then feature nodes are extracted on top of the components. An elastic net regularization technique is used to constrain the output weights of the nodes, preserving the significant nodes, and finally obtaining a more sparse MSBLS. Experiments are conducted using the NASA publicly available commercial modular aero-propulsion system simulation (C-MAPSS) dataset and the N-CMAPSS dataset, and our proposed MSBLS not only improves the accuracy of RUL prediction but also has a very short training time compared with the latest research methods nowadays.

Underwater image enhancement (UIE) is an essential task for intelligent environment perception in underwater remote visual sensing scenarios. However, the computing power of mobile platforms limits the usage of larger-scale models. In this paper, we propose a lightweight encoder-decoder architecture (UIENet) to enhance underwater images from visual sensors. We also involve the architecture into a generative adversarial model (UIEGAN) against a supervised discriminator to further perfect its corrective capabilities for the photo-realistic images with more global appearance and local details. The multi-resolution counterparts are embedded into the generator to diversify the feature representation of the original inputs. Further, UIEGAN guides the spatial attention module and the channel attention module to jointly enhance the global-local connection of the image. We evaluate the proposed method on benchmark datasets of UIEB and UFO-120 and report better performance than the state-of-the-art schemes, exceeding 11.15% and 12.85% on peak signal-to-noise ratio (PSNR) than the baselines of these datasets. Besides, by testing on the UIEB challenge, URPC and SQUID datasets without any reference images, our scheme outperforms the other methods on evaluation metrics to validate its generalization performance, and meanwhile uses a series of ablation study demonstrates the effectiveness of the functional modules.

Specific emitter identification (SEI) is a technology that extracts subtle features from signals sent by emitters to identify different individuals. It can effectively improve the security of the Industrial Internet of Things (IIoT) by acting on the physical layer of the internet. Recent research on SEI has focused on deep learning (DL) models that can automatically learn effective inherent emitter features from raw signals. Nevertheless, training popular DL models is computationally expensive because of the numerous hyperparameters and nonscalable structures. This limits the application of DL-based SEI models in certain practical IIoT scenarios. To address this concern, we propose an adaptive broad learning (ABL) method to build a lightweight SEI model. In the proposed model, the raw signal samples are mapped to feature nodes, and the emitters are denoted as the output nodes. The hidden nodes are directly connected to the output nodes by a broad network. Through this flat structure, the size and calculation amount of the model can be effectively reduced. To further economize the computational cost, we designed an adaptive node expansion strategy for rapidly obtaining the optimal hyperparameters of the models. The results of experiments on real-world data prove the superiority of ABL over popular state-of-the-art DL-based SEI models.

With the rapid development of edge intelligence (EI) and machine learning (ML), the applications of Cyber-Physical Systems (CPS) have been discovered in all aspects of the life world. As one of its most essential branches, Medical CPS (MCPS) determines human health and medical treatment in the Internet of Everything (IOE) era. Knowledge sharing is the critical point of MCPS and has also been humanity's best dream through the ages. This paper explores a novel knowledge-sharing model in MCPS and takes a pulmonary nodule detection task as a significant case for building an Unet-based mask generator. A Classification-guided Module (CGM)-based discriminator with knowledge from EMRs is set against a generator to offer a promising result for each mask from the inexperienced participant of federated ML. After an iterative communication between the federated server and its clients for knowledge sharing, the segmented sub-image owns a coincident attribute distribution with that of the EMRs from the experts. Besides, the adversarial network augment the data to normalize the data distribution for all the clients as a remission for none independent identically distributed (non-IID) data problem. We implement a detection framework on the simulated EI environment following an existing adaptive synchronization strategy based on data sharing and median loss function. On 1304 scans of the merged dataset, our proposed framework can help boost the detection performance for most of the existing methods of pulmonary nodule detection.

In recent years, the increasing number of industrial infrastructure security incidents around the world has drawn public attention to industrial control networks (ICNs) security issues. Fault diagnosis of industrial devices is an indispensable part of the security system in ICNs. The mainstream fault diagnosis models rely on long-term training and massive fault data, which results in the inability to update the model effectively and timely when the environment changes. Thus, some researchers focus on developing cross-domain industrial fault diagnosis methods. However, they usually presume that the samples of the target and source domains share the same fault mode sets, and existing prior knowledge concerning the label spaces of these two domains. These are difficult to satisfy in actual ICNs. To respond to these challenges, we develop atransferability-measured adversarial adaptation network(TAAN) to identify unknown classes without prior knowledge. It embeds the hybrid transferability estimation into an adversarial domain adaptive network to weigh the contribution of each sample. In this way, TAAN can properly classify samples in a public label space by selectively aligning source and target samples with high transferability. The experimental results obtained using two diagnosis datasets prove that the developed TAAN can achieve satisfactory diagnostic accuracy by effectively bridging the distribution discrepancy under various working conditions.

Security assessment (SA) system is crucial to ensure the production safety of a smart factory with rapid development of artificial intelligence. In this article, we propose a novel SA framework. Different from the conventional static monitoring systems based on traditional sensing technologies, the proposed framework can automatically detect objects via visual sensing. We use a skeleton-based graph convolutional network to generate action vocabulary for the intermediate representations of action-to-action cooccurrence relations. These representations are encoded into the sequential interaction models to form the interaction representations. Integrating the states of molten steel levels as the reference labels, the sequential representations are fed into a recurrent neural network model with multilayer gated recurrent units (GRUs) to capture the key interactions leading to the accidents, in which an attention mechanism is used to reweight the actions and eliminate the invalid interactions. The predicted labels and the hidden states of the scenes are passing among multilayer GRUs. Finally, we optimize the global output to dynamically assess the security by calculating a joint objective function with a regularized cross-entropy loss. On the self-collected dataset from our partner Iron and Steel company and on-line video clips, the proposed framework performs better than the existing SA schemes.

Depression and anxiety are a couple of pernicious mental states, which may affect lifestyle and quality and even become the primary causes of disability worldwide. Hence, daily monitoring of the mental states is significant for avoiding possible injury. Dynamics of the human blood vascular system convey significant information on recording the emotion and the mental state, which can be monitored via photoplethysmography (PPG). It is one of the best schemes with the advantages of nonintrusiveness and low cost. Conventional approaches for PPG signal analysis usually depend on handcrafted feature extraction and classification, thus resulting in a lack of feature discrimination and difficulties in generalization. In this article, we propose an attentive deep recurrent architecture called Arousal-valence Networks (ArvaNets), which benefits from graph convolutional networks and recurrent neural networks. Our approach overcomes the limitations of previous methods by automatically extracting the learnable spatial representations from a rigorous custom data set as semantic motifs to infer immediate emotions, which are mapped to a 2-D arousal-valence coordinate system. Finally, we exploit long short-term memory (LSTM) units to output the mental states by incorporating the temporal factor. During the entire inference, we propose a spatiotemporal attention mechanism based on correlation fractal dimensions (CFDs) and time-averaged wall shear stress (TAWSS) to capture and stress the key subtle motifs for performance optimization. Experimental results demonstrate the proposed architecture has enough competitiveness in the tasks of emotion and mental-state recognition for daily monitoring.

With the development of the industrial Internet of Things and the advancements in wireless sensor networking technologies, the smart grid based on edge computing now is regarded as being essential for real-time monitoring and automatic control of the electricity generation and distribution. In this article, we propose a highly efficient energy control framework supported by edge computing to reduce energy waste and increase the benefit for industrial users. To this end, battery energy storage systems (BESSs) are currently being employed to store energy for stability of supply and quality of power. The optimal load patterns and corresponding energy storage capacities of the BESSs can be obtained through the framework, according to the energy market and the historical load data of industrial users. However, computing these requires considering the tradeoff between equipment cost, time-of-use electricity price, running expenses, and other related factors, which would be an NP-hard problem. To address this challenge, we also propose an adaptive mixed differential evolution algorithm with a novel mutation strategy. Experiments on real-world data demonstrate the effectiveness of the proposed algorithm and framework.

This paper proposes a two-way multi-ringed forest (TMR-Forest) to estimating the malignancy of the pulmonary nodules for false positive reduction (FPR). Based on our previous work of deep decision framework, named MR-Forest, we generate a growing path mode on predefined pseudo-timeline of L time slots to build pseudo-spatiotemporal features. It synchronously works with FPR based on MR-Forest to help predict the labels from a dynamic perspective. Concretely, Mask R-CNN is first used to recommend the bounding boxes of ROIs and classify their pathological features. Afterward, hierarchical attribute matching is introduced to obtain the input ROIs' attribute layouts and select the candidates for their growing path generation. The selected ROIs can replace the fixed-sized ROIs' fitting results at different time slots for data augmentation. A two-stage counterfactual path elimination is used to screen out the input paths of the cascade forest. Finally, a simple label selection strategy is executed to output the predicted label to point out the input nodule's malignancy. On 1034 scans of the merged dataset, the framework can report more accurate malignancy labels to achieve a better CPM score of 0.912, which exceeds those of MR-Forest and 3DDCNNs about 2.8% and 4.7%, respectively.

With the development of deep learning methods such as convolutional neural network (CNN), the accuracy of automated pulmonary nodule detection has been greatly improved. However, the high computational and storage costs of the large-scale network have been a potential concern for the future widespread clinical application. In this paper, an alternative Multi-ringed (MR)-Forest framework, against the resource-consuming neural networks (NN)-based architectures, has been proposed for false positive reduction in pulmonary nodule detection, which consists of three steps. First, a novel multi-ringed scanning method is used to extract the order ring facets (ORFs) from the surface voxels of the volumetric nodule models; Second, Mesh-LBP and mapping deformation are employed to estimate the texture and shape features. By sliding and resampling the multi-ringed ORFs, feature volumes with different lengths are generated. Finally, the outputs of multi-level are cascaded to predict the candidate class. On 1034 scans merging the dataset from the Affiliated Hospital of Liaoning University of Traditional Chinese Medicine (AH-LUTCM) and the LUNA16 Challenge dataset, our framework performs enough competitiveness than state-of-the-art in false positive reduction task (CPM score of 0.865). Experimental results demonstrate that MR-Forest is a successful solution to satisfy both resource-consuming and effectiveness for automated pulmonary nodule detection. The proposed MR-forest is a general architecture for 3D target detection, it can be easily extended in many other medical imaging analysis tasks, where the growth trend of the targeting object is approximated as a spheroidal expansion.