IEEE/CAA Journal of Automatica Sinica

Parallel Seeds: From Foundation Models to Foundation Intelligence for Agricultural Sustainability

Laiyi Fu, Shunkang Ling, Danyang Wu, Mengzhen Kang, Fei-Yue Wang, Hequan Sun

2025, 12(3): 481-484. doi: 10.1109/JAS.2024.124914

Abstract(13) HTML (3) PDF(3)

Abstract:

When Embodied AI Meets Industry 5.0: Human-Centered Smart Manufacturing

Jing Xu, Qiyu Sun, Qing-Long Han, Yang Tang

2025, 12(3): 485-501. doi: 10.1109/JAS.2025.125327

Abstract(46) HTML (5) PDF(12)

Abstract:
As embodied intelligence (EI), large language models (LLMs), and cloud computing continue to advance, Industry 5.0 facilitates the development of industrial artificial intelligence (IndAI) through cyber-physical-social systems (CPSSs) with a human-centric focus. These technologies are organized by the system-wide approach of Industry 5.0, in order to empower the manufacturing industry to achieve broader societal goals of job creation, economic growth, and green production. This survey first provides a general framework of smart manufacturing in the context of Industry 5.0. Wherein, the embodied agents, like robots, sensors, and actuators, are the carriers for IndAI, facilitating the development of the self-learning intelligence in individual entities, the collaborative intelligence in production lines and factories (smart systems), and the swarm intelligence within industrial clusters (systems of smart systems). Through the framework of CPSSs, the key technologies and their possible applications for supporting the single-agent, multi-agent and swarm-agent embodied IndAI have been reviewed, such as the embodied perception, interaction, scheduling, multi-mode large language models, and collaborative training. Finally, to stimulate future research in this area, the open challenges and opportunities of applying Industry 5.0 to smart manufacturing are identified and discussed. The perspective of Industry 5.0-driven manufacturing industry aims to enhance operational productivity and efficiency by seamlessly integrating the virtual and physical worlds in a human-centered manner, thereby fostering an intelligent, sustainable, and resilient industrial landscape.

PromptFusion: Harmonized Semantic Prompt Learning for Infrared and Visible Image Fusion

Jinyuan Liu, Xingyuan Li, Zirui Wang, Zhiying Jiang, Wei Zhong, Wei Fan, Bin Xu

2025, 12(3): 502-515. doi: 10.1109/JAS.2024.124878

Abstract(218) HTML (6) PDF(62)

Abstract:
The goal of infrared and visible image fusion (IVIF) is to integrate the unique advantages of both modalities to achieve a more comprehensive understanding of a scene. However, existing methods struggle to effectively handle modal disparities, resulting in visual degradation of the details and prominent targets of the fused images. To address these challenges, we introduce PromptFusion, a prompt-based approach that harmoniously combines multi-modality images under the guidance of semantic prompts. Firstly, to better characterize the features of different modalities, a contourlet autoencoder is designed to separate and extract the high-/low-frequency components of different modalities, thereby improving the extraction of fine details and textures. We also introduce a prompt learning mechanism using positive and negative prompts, leveraging Vision-Language Models to improve the fusion model’s understanding and identification of targets in multi-modality images, leading to improved performance in downstream tasks. Furthermore, we employ bi-level asymptotic convergence optimization. This approach simplifies the intricate non-singleton non-convex bi-level problem into a series of convergent and differentiable single optimization problems that can be effectively resolved through gradient descent. Our approach advances the state-of-the-art, delivering superior fusion quality and boosting the performance of related downstream tasks. Project page: https://github.com/hey-it-s-me/PromptFusion.

Nonlinear Control for Unstable Networked Plants in the Presence of Actuator and Sensor Limitations Using Robust Right Coprime Factorization

Yuanhong Xu, Mingcong Deng

2025, 12(3): 516-527. doi: 10.1109/JAS.2024.124854

Abstract(67) HTML (2) PDF(16)

Abstract:
In this paper, a nonlinear control approach for an unstable networked plant in the presence of actuator and sensor limitations using robust right coprime factorization is proposed. The actuator is limited by upper and lower constraints and the sensor in the feedback loop is subjected to network-induced unknown time-varying delay and noise. With this nonlinear control method, we first employ right coprime factorization based on isomorphism and operator theory to factorize the plant, so that bounded input bounded output (BIBO) stability can be guaranteed. Next, continuous-time generalized predictive control (CGPC) is utilized for the unstable operator of the right coprime factorized plant to guarantee inner stability and enables the closed-loop dynamics of the system with predictive characteristics. Meanwhile, a second-DoF (degrees of freedom) switched controller that satisfies a perturbed Bezout identity and a robustness condition is designed. By using the CGPC controller that possesses predictive behavior and the second-DoF switched stabilizer, the overall stability of the plant subjected to actuator limitations is guaranteed. To address sensor limitations that exist in networked plants in the form of delay and noise which often cause system performance degradation, we implement an identity operator definition in the feedback loop to compensate for these adverse effects. Further, a pre-operator is designed to ensure that the plant output tracks the reference input. Finally, the effectiveness of the proposed design scheme is demonstrated by simulations.

Multi-Scale Time Series Segmentation Network Based on Eddy Current Testing for Detecting Surface Metal Defects

Xiaorui Li, Xiaojuan Ban, Haoran Qiao, Zhaolin Yuan, Hong-Ning Dai, Chao Yao, Yu Guo, Mohammad S. Obaidat, George Q. Huang

2025, 12(3): 528-538. doi: 10.1109/JAS.2025.125117

Abstract(78) HTML (3) PDF(24)

Abstract:
In high-risk industrial environments like nuclear power plants, precise defect identification and localization are essential for maintaining production stability and safety. However, the complexity of such a harsh environment leads to significant variations in the shape and size of the defects. To address this challenge, we propose the multivariate time series segmentation network (MSSN), which adopts a multiscale convolutional network with multi-stage and depth-separable convolutions for efficient feature extraction through variable-length templates. To tackle the classification difficulty caused by structural signal variance, MSSN employs logarithmic normalization to adjust instance distributions. Furthermore, it integrates classification with smoothing loss functions to accurately identify defect segments amid similar structural and defect signal subsequences. Our algorithm evaluated on both the Mackey-Glass dataset and industrial dataset achieves over 95% localization and demonstrates the capture capability on the synthetic dataset. In a nuclear plant’s heat transfer tube dataset, it captures 90% of defect instances with 75% middle localization F1 score.

Fault Warning of Satellite Momentum Wheels With a Lightweight Transformer Improved by FastDTW

Yiming Gao, Shi Qiu, Ming Liu, Lixian Zhang, Xibin Cao

2025, 12(3): 539-549. doi: 10.1109/JAS.2024.124689

Abstract(81) HTML (4) PDF(18)

Abstract:
The momentum wheel assumes a dominant role as an inertial actuator for satellite attitude control systems. Due to the effects of structural aging and external interference, the momentum wheel may experience the gradual emergence of irreversible faults. These fault features will become apparent in the telemetry signal transmitted by the momentum wheel. This paper introduces ADTWformer, a lightweight model for long-term prediction of time series, to analyze the time evolution trend and multi-dimensional data coupling mechanism of satellite momentum wheel faults. Moreover, the incorporation of the approximate Markov blanket with the maximum information coefficient presents a novel methodology for performing correlation analysis, providing significant perspectives from a data-centric standpoint. Ultimately, the creation of an adaptive alarm mechanism allows for the successful attainment of the momentum wheel fault warning by detecting the changes in the health status curves. The analysis methodology outlined in this article has exhibited positive results in identifying instances of satellite momentum wheel failure in two scenarios, thereby showcasing considerable promise for large-scale applications.

Optimal Production Capacity Matching for Blockchain-Enabled Manufacturing Collaboration With the Iterative Double Auction Method

Ying Chen, Feilong Lin, Zhongyu Chen, Changbing Tang, Cailian Chen

2025, 12(3): 550-562. doi: 10.1109/JAS.2024.124626

Abstract(50) HTML (4) PDF(9)

Abstract:
The increased demand for personalized customization calls for new production modes to enhance collaborations among a wide range of manufacturing practitioners who unnecessarily trust each other. In this article, a blockchain-enabled manufacturing collaboration framework is proposed, with a focus on the production capacity matching problem for blockchain-based peer-to-peer (P2P) collaboration. First, a digital model of production capacity description is built for trustworthy and transparent sharing over the blockchain. Second, an optimization problem is formulated for P2P production capacity matching with objectives to maximize both social welfare and individual benefits of all participants. Third, a feasible solution based on an iterative double auction mechanism is designed to determine the optimal price and quantity for production capacity matching with a lack of personal information. It facilitates automation of the matching process while protecting users’ privacy via blockchain-based smart contracts. Finally, simulation results from the Hyperledger Fabric-based prototype show that the proposed approach increases social welfare by 1.4% compared to the Bayesian game-based approach, makes all participants profitable, and achieves 90% fairness of enterprises.

Dynamic Process Monitoring Based on Dot Product Feature Analysis for Thermal Power Plants

Xin Ma, Tao Chen, Youqing Wang

2025, 12(3): 563-574. doi: 10.1109/JAS.2024.124908

Abstract(57) HTML (4) PDF(26)

Abstract:
Data-driven process monitoring is an effective approach to assure safe operation of modern manufacturing and energy systems, such as thermal power plants being studied in this work. Industrial processes are inherently dynamic and need to be monitored using dynamic algorithms. Mainstream dynamic algorithms rely on concatenating current measurement with past data. This work proposes a new, alternative dynamic process monitoring algorithm, using dot product feature analysis (DPFA). DPFA computes the dot product of consecutive samples, thus naturally capturing the process dynamics through temporal correlation. At the same time, DPFA’s online computational complexity is lower than not just existing dynamic algorithms, but also classical static algorithms (e.g., principal component analysis and slow feature analysis). The detectability of the new algorithm is analyzed for three types of faults typically seen in process systems: sensor bias, process fault and gain change fault. Through experiments with a numerical example and real data from a thermal power plant, the DPFA algorithm is shown to be superior to the state-of-the-art methods, in terms of better monitoring performance (fault detection rate and false alarm rate) and lower computational complexity.

Consensus Control Strategy for the Treatment of Tumour With Neuroadaptive Cellular Immunotherapy

Jiayue Sun, Dongni Li, Huaguang Zhang, Lu Liu, Wenyue Zhao

2025, 12(3): 575-584. doi: 10.1109/JAS.2024.124941

Abstract(61) HTML (4) PDF(9)

Abstract:
This paper presents a novel neuro-adaptive cellular immunotherapy control strategy that leverages the high efficiency and applicability of chimeric antigen receptor-engineered T (CAR-T) cells in treating cancer. The proposed real-time control strategy aims to maximize tumor regression while ensuring the safety of the treatment. A dynamic growth model of cancer cells under the influence of cellular immunotherapy is established for the first time, which aligns with clinical experimental results. Utilizing the backstepping method, a novel consensus reference model is designed to consider the characteristics of cancer cell changes during the treatment process and conform to clinical rules. The model is segmented and continuous, with cancer cells expected to decrease in a step-like manner. Furthermore, a prescribed performance mechanism is constructed to maintain the therapeutic effect of the proposed scheme while ensuring the transient performance of the system. Through the analysis of Lyapunov stability, all signals within the closed-loop system are proven to be semiglobally uniformly ultimately bounded (SGUUB). Simulation results demonstrate the effectiveness of the proposed control strategy, highlighting its potential for clinical application in cancer treatment.

Cyber-Attacks With Resource Constraints on Discrete Event Systems Under Supervisory Control

Zhaoyang He, Naiqi Wu, Rong Su, Zhiwu Li

2025, 12(3): 585-595. doi: 10.1109/JAS.2024.124596

Abstract(101) HTML (3) PDF(31)

Abstract:
With the development of cyber-physical systems, system security faces more risks from cyber-attacks. In this work, we study the problem that an external attacker implements covert sensor and actuator attacks with resource constraints (the total resource consumption of the attacks is not greater than a given initial resource of the attacker) to mislead a discrete event system under supervisory control to reach unsafe states. We consider that the attacker can implement two types of attacks: One by modifying the sensor readings observed by a supervisor and the other by enabling the actuator commands disabled by the supervisor. Each attack has its corresponding resource consumption and remains covert. To solve this problem, we first introduce a notion of combined-attackability to determine whether a closed-loop system may reach an unsafe state after receiving attacks with resource constraints. We develop an algorithm to construct a corrupted supervisor under attacks, provide a verification method for combined-attackability in polynomial time based on a plant, a corrupted supervisor, and an attacker’s initial resource, and propose a corresponding attack synthesis algorithm. The effectiveness of the proposed method is illustrated by an example.

Enhanced Tube-Based Event-Triggered Stochastic Model Predictive Control With Additive Uncertainties

Chenxi Gu, Xinli Wang, Kang Li, Xiaohong Yin, Shaoyuan Li, Lei Wang

2025, 12(3): 596-605. doi: 10.1109/JAS.2024.124974

Abstract(68) HTML (2) PDF(20)

Abstract:
This paper proposes an event-triggered stochastic model predictive control for discrete-time linear time-invariant (LTI) systems under additive stochastic disturbances. It first constructs a probabilistic invariant set and a probabilistic reachable set based on the priori knowledge of system uncertainties. Assisted with enhanced robust tubes, the chance constraints are then formulated into a deterministic form. To alleviate the online computational burden, a novel event-triggered stochastic model predictive control is developed, where the triggering condition is designed based on the past and future optimal trajectory tracking errors in order to achieve a good trade-off between system resource utilization and control performance. Two triggering parameters σ and γ are used to adjust the frequency of solving the optimization problem. The probabilistic feasibility and stability of the system under the event-triggered mechanism are also examined. Finally, numerical studies on the control of a heating, ventilation, and air conditioning (HVAC) system confirm the efficacy of the proposed control.

A Diffusion Model for Traffic Data Imputation

Bo Lu, Qinghai Miao, Yahui Liu, Tariku Sinshaw Tamir, Hongxia Zhao, Xiqiao Zhang, Yisheng Lv, Fei-Yue Wang

2025, 12(3): 606-617. doi: 10.1109/JAS.2024.124611

Abstract(92) HTML (3) PDF(28)

Abstract:
Imputation of missing data has long been an important topic and an essential application for intelligent transportation systems (ITS) in the real world. As a state-of-the-art generative model, the diffusion model has proven highly successful in image generation, speech generation, time series modelling etc. and now opens a new avenue for traffic data imputation. In this paper, we propose a conditional diffusion model, called the implicit-explicit diffusion model, for traffic data imputation. This model exploits both the implicit and explicit feature of the data simultaneously. More specifically, we design two types of feature extraction modules, one to capture the implicit dependencies hidden in the raw data at multiple time scales and the other to obtain the long-term temporal dependencies of the time series. This approach not only inherits the advantages of the diffusion model for estimating missing data, but also takes into account the multiscale correlation inherent in traffic data. To illustrate the performance of the model, extensive experiments are conducted on three real-world time series datasets using different missing rates. The experimental results demonstrate that the model improves imputation accuracy and generalization capability.

Spiking Reinforcement Learning Enhanced by Bioinspired Event Source of Multi-dendrite Spiking Neuron and Dynamic Thresholds

Xingyue Liang, Qiaoyun Wu, Yun Zhou, Chunyu Tan, Hongfu Yin, Changyin Sun

2025, 12(3): 618-629. doi: 10.1109/JAS.2024.124551

Abstract(144) HTML (2) PDF(36)

Abstract:
Deep reinforcement learning (DRL) achieves success through the representational capabilities of deep neural networks (DNNs). Compared to DNNs, spiking neural networks (SNNs), known for their binary spike information processing, exhibit more biological characteristics. However, the challenge of using SNNs to simulate more biologically characteristic neuronal dynamics to optimize decision-making tasks remains, directly related to the information integration and transmission in SNNs. Inspired by the advanced computational power of dendrites in biological neurons, we propose a multi-dendrite spiking neuron (MDSN) model based on Multi-compartment spiking neurons (MCN), expanding dendrite types from two to multiple and deriving the analytical solution of somatic membrane potential. We apply the MDSN to deep distributional reinforcement learning to enhance its performance in executing complex decision-making tasks. The proposed model can effectively and adaptively integrate and transmit meaningful information from different sources. Our model uses a bioinspired event-enhanced dendrite structure to emphasize features. Meanwhile, by utilizing dynamic membrane potential thresholds, it adaptively maintains the homeostasis of MDSN. Extensive experiments on Atari games show that the proposed model outperforms some state-of-the-art spiking distributional RL models by a significant margin.