IEEE/CAA Journal of Automatica Sinica

Parallel Medical Devices and Instruments: Integrating Edge and Cloud Intelligence for Smart Treatment and Health Systems

Fei Lin, Tommy Gao, Dali Sun, Qinghua Ni, Xianting Ding, Jing Wang, David Wenzhong Gao, Fei-Yue Wang

2025, 12(4): 651-654. doi: 10.1109/JAS.2024.124614

Abstract(155) HTML (30) PDF(113)

Abstract:

Length-Variable Bionic Continuum Robot With Millimeter-Scale Diameter and Compliant Driving Force

Shiying Zhao, Qingxin Meng, Xuzhi Lai, Jinhua She, Edwardo Fumihiko Fukushima, Min Wu

2025, 12(4): 655-667. doi: 10.1109/JAS.2024.125091

Abstract(310) HTML (32) PDF(76)

Abstract:
Compared with conventional rigid-link robots, bionic continuum robots (CRs) show great potential in unstructured environments because of their adaptivity and continuous deformation ability. However, designing a CR to achieve miniaturization, variable length and compliant driving force remains a challenge. Here, inspired by the earthworm in nature, we report a length-variable bionic CR with millimeter-scale diameter and compliant driving force. The CR consists of two main components: the robot body and soft drives. The robot body is only 6 mm in diameter, and is composed of a backbone and transmission devices. The backbone is divided into three segments, and each segment is capable of adjusting its length and bending like the earthworm. The maximum length variation of the backbone can reach an astonishing 70 mm with a backbone’s initial length of 150 mm, and the maximum bending angle of each segment can reach 120 degrees. In addition, we develop soft drives using pneumatic soft actuators (PSAs) as a replacement for the rigid motors typically used in conventional CRs. These soft drives control the motions of the transmission devices, enabling length variation and bending of the backbone. By utilizing these soft drives, we ensure that the robot body has a compliant driving force, which addresses users’ concerns about human safety during interactions. In practical applications, we prove that this CR can perform delicate manipulations by successfully completing writing tasks. Additionally, we show its application value for detections and medical treatments by entering the narrow tube and the oral.

A Self-Healing Predictive Control Method for Discrete-Time Nonlinear Systems

Shulei Zhang, Runda Jia

2025, 12(4): 668-682. doi: 10.1109/JAS.2024.124620

Abstract(184) HTML (27) PDF(61)

Abstract:
In this work, a self-healing predictive control method for discrete-time nonlinear systems is presented to ensure the system can be safely operated under abnormal states. First, a robust MPC controller for the normal case is constructed, which can drive the system to the equilibrium point when the closed-loop states are in the predetermined safe set. In this controller, the tubes are built based on the incremental Lyapunov function to tighten nominal constraints. To deal with the infeasible controller when abnormal states occur, a self-healing predictive control method is further proposed to realize self-healing by driving the system towards the safe set. This is achieved by an auxiliary soft-constrained recovery mechanism that can solve the constraint violation caused by the abnormal states. By extending the discrete-time robust control barrier function theory, it is proven that the auxiliary problem provides a predictive control barrier bounded function to make the system asymptotically stable towards the safe set. The theoretical properties of robust recursive feasibility and bounded stability are further analyzed. The efficiency of the proposed controller is verified by a numerical simulation of a continuous stirred-tank reactor process.

Accumulative-Error-Based Event-Triggered Control for Discrete-Time Linear Systems: A Discrete-Time Looped Functional Method

Xian-Ming Zhang, Qing-Long Han, Xiaohua Ge, Bao-Lin Zhang

2025, 12(4): 683-693. doi: 10.1109/JAS.2024.124476

Abstract(353) HTML (26) PDF(57)

Abstract:
This paper is concerned with event-triggered control of discrete-time systems with or without input saturation. First, an accumulative-error-based event-triggered scheme is devised for control updates. When the accumulated error between the current state and the latest control update exceeds a certain threshold, an event is triggered. Such a scheme can ensure the event-generator works at a relatively low rate rather than falls into hibernation especially after the system steps into its steady state. Second, the looped functional method for continuous-time systems is extended to discrete-time systems. By introducing an innovative looped functional that links the event-triggered scheme, some sufficient conditions for the co-design of control gain and event-triggered parameters are obtained in terms of linear matrix inequalities with a couple of tuning parameters. Then, the proposed method is applied to discrete-time systems with input saturation. As a result, both suitable control gains and event-triggered parameters are also co-designed to ensure the system trajectories converge to the region of attraction. Finally, an unstable reactor system and an inverted pendulum system are given to show the effectiveness of the proposed method.

Necessary and Sufficient Conditions for Controllability and Essential Controllability of Directed Circle and Tree Graphs

Jijun Qu, Zhijian Ji, Jirong Wang, Yungang Liu

2025, 12(4): 694-704. doi: 10.1109/JAS.2024.124866

Abstract(131) HTML (29) PDF(40)

Abstract:
The multi-agent controllability is intrinsically affected by the network topology and the selection of leaders. A focus of exploring this problem is to uncover the relationship between the eigenspace of Laplacian matrix and network topology. For strongly connected directed circle graphs, we elaborate how the zero entries in the left eigenvectors of Laplacian matrix L arise. The topologies arising from left eigenvectors with zero entries are filtered to construct essentially controllable directed circle graphs regardless of the choice of leaders. We propose two methods for constructing a substantial quantity of essentially controllable graphs, with a focus on utilizing essentially controllable circle graphs as the foundation. For a special directed graph-OT tree, the controllability is shown to be related with its substructure-paths. This promotes the establishment of a sufficient and necessary condition for controllability. Finally, a method is presented to check the controllable subspace by identifying the left eigenvectors and generalized left eigenvectors.

Dissecting and Mitigating Semantic Discrepancy in Stable Diffusion for Image-to-Image Translation

Yifan Yuan, Guanqun Yang, James Z. Wang, Hui Zhang, Hongming Shan, Fei-Yue Wang, Junping Zhang

2025, 12(4): 705-718. doi: 10.1109/JAS.2024.124800

Abstract(313) HTML (31) PDF(66)

Abstract:
Finding suitable initial noise that retains the original image’s information is crucial for image-to-image (I2I) translation using text-to-image (T2I) diffusion models. A common approach is to add random noise directly to the original image, as in SDEdit. However, we have observed that this can result in “semantic discrepancy” issues, wherein T2I diffusion models misinterpret the semantic relationships and generate content not present in the original image. We identify that the noise introduced by SDEdit disrupts the semantic integrity of the image, leading to unintended associations between unrelated regions after U-Net upsampling. Building on the widely-used latent diffusion model, Stable Diffusion, we propose a training-free, plug-and-play method to alleviate semantic discrepancy and enhance the fidelity of the translated image. By leveraging the deterministic nature of denoising diffusion implicit models (DDIMs) inversion, we correct the erroneous features and correlations from the original generative process with accurate ones from DDIM inversion. This approach alleviates semantic discrepancy and surpasses recent DDIM-inversion-based methods such as PnP with fewer priors, achieving a speedup of 11.2 times in experiments conducted on COCO, ImageNet, and ImageNet-R datasets across multiple I2I translation tasks. The codes are available at https://github.com/Sherlockyyf/Semantic_Discrepancy.

Variable Reconstruction for Evolutionary Expensive Large-Scale Multiobjective Optimization and Its Application on Aerodynamic Design

Jianqing Lin, Cheng He, Ye Tian, Linqiang Pan

2025, 12(4): 719-733. doi: 10.1109/JAS.2024.124947

Abstract(117) HTML (29) PDF(36)

Abstract:
Expensive multiobjective optimization problems (EMOPs) are complex optimization problems exacted from real-world applications, where each objective function evaluation (FE) involves expensive computations or physical experiments. Many surrogate-assisted evolutionary algorithms (SAEAs) have been designed to solve EMOPs. Nevertheless, EMOPs with large-scale decision variables remain challenging for existing SAEAs, leading to difficulties in maintaining convergence and diversity. To address this deficiency, we proposed a variable reconstruction-based SAEA (VREA) to balance convergence enhancement and diversity maintenance. Generally, a cluster-based variable reconstruction strategy reconstructs the original large-scale decision variables into low-dimensional weight variables. Thus, the population can be rapidly pushed towards the Pareto set (PS) by optimizing low-dimensional weight variables with the assistance of surrogate models. Population diversity is improved due to the cluster-based variable reconstruction strategy. An adaptive search step size strategy is proposed to balance exploration and exploitation further. Experimental comparisons with four state-of-the-art SAEAs are conducted on benchmark EMOPs with up to 1000 decision variables and an aerodynamic design task. Experimental results demonstrate that VREA obtains well-converged and diverse solutions with limited real FEs.

A k-Winners-Take-All (kWTA) Network With Noise Characteristics Captured

Jiexing Li, Yulin Cao, Zhengtai Xie, Long Jin

2025, 12(4): 734-744. doi: 10.1109/JAS.2025.125153

Abstract(131) HTML (28) PDF(34)

Abstract:
Competition-based $ k $-winners-take-all ($ k $WTA) networks play a crucial role in multi-agent systems. However, existing $ k $WTA networks either neglect the impact of noise or only consider simple forms, such as constant noise. In practice, noises often exhibit time-varying and nonlinear characteristics, which can be modeled using nonlinear functions and approximated by high-order polynomials. Such noises pose significant challenges for current $ k $WTA networks, limiting their practical applications. To address this, a $ k $WTA network with noise characteristics captured ($ k $WTA-NCC) is proposed in this article. Theoretical analyses demonstrate that the residual error of the proposed $ k $WTA-NCC network converges to zero globally, while simulation results confirm its robustness against polynomial noises. Additionally, a $ k $WTA coordination model is constructed by integrating the proposed network with a consensus estimator to achieve multi-agent tracking tasks. Finally, simulations and physical experiments are conducted further to demonstrate the validity and practicality of the $ k $WTA coordination model.

E²AG: Entropy-Regularized Ensemble Adaptive Graph for Industrial Soft Sensor Modeling

Zhichao Chen, Licheng Pan, Yiran Ma, Zeyu Yang, Le Yao, Jinchuan Qian, Zhihuan Song

2025, 12(4): 745-760. doi: 10.1109/JAS.2024.124884

Abstract(142) HTML (39) PDF(26)

Abstract:
Adaptive graph neural networks (AGNNs) have achieved remarkable success in industrial process soft sensing by incorporating explicit features that delineate the relationships between process variables. This article introduces a novel GNN framework, termed entropy-regularized ensemble adaptive graph (E²AG), aimed at enhancing the predictive accuracy of AGNNs. Specifically, this work pioneers a novel AGNN learning approach based on mirror descent, which is central to ensuring the efficiency of the training procedure and consequently guarantees that the learned graph naturally adheres to the row-normalization requirement intrinsic to the message-passing of GNNs. Subsequently, motivated by multi-head self-attention mechanism, the training of ensembled AGNNs is rigorously examined within this framework, incorporating an entropy regularization term in the learning objective to ensure the diversity of the learned graph. After that, the architecture and training algorithm of the model are then concisely summarized. Finally, to ascertain the efficacy of the proposed E²AG model, extensive experiments are conducted on real-world industrial datasets. The evaluation focuses on prediction accuracy, model efficacy, and sensitivity analysis, demonstrating the superiority of E²AG in industrial soft sensing applications.

Collision-Free Maneuvering for a UAV Swarm Based on Parallel Control

Jiacheng Li, Wenhui Ma, YangWang Fang, Dengxiu Yu, C. L. Philip Chen

2025, 12(4): 761-775. doi: 10.1109/JAS.2024.124674

Abstract(113) HTML (26) PDF(37)

Abstract:
The maneuvering of a large-scale unmanned aerial vehicle (UAV) swarm, notable for flexible flight with collision-free, is still challenging due to the significant number of UAVs and the compact configuration of the swarm. In light of this problem, a novel parallel control method that utilizes space and time transformation is proposed. First, the swarm is decomposed based on a grouping-hierarchical strategy, while the distinct flight roles are assigned to each UAV. Then, to achieve the desired configuration (DCF) in the real world, a bijection transformation is conducted in the space domain, converting an arbitrarily general configuration (GCF) into a standard configuration (SCF) in the virtual space. Further, to improve the flexibility of the swarm, the time scaling transformation is adopted in the time domain, which ensures the desired prescribed-time convergence of the swarm independent of initial conditions. Finally, simulation results demonstrate that collision-free maneuvering, including formation changes and turning, can be effectively and rapidly achieved by the proposed parallel control method. Overall, this research contributes a viable solution for enhancing cooperation among large-scale UAV swarms.

Multi-Spacecraft Formation Control Under False Data Injection Attack: A Cross Layer Fuzzy Game Approach

Yifan Zhong, Yuan Yuan, Huanhuan Yuan, Mengbi Wang, Huaping Liu

2025, 12(4): 776-788. doi: 10.1109/JAS.2024.124872

Abstract(203) HTML (25) PDF(64)

Abstract:
In this paper, we address a cross-layer resilient control issue for a kind of multi-spacecraft system (MSS) under attack. Attackers with bad intentions use the false data injection (FDI) attack to prevent the MSS from reaching the goal of consensus. In order to ensure the effectiveness of the control, the embedded defender in MSS preliminarily allocates the defense resources among spacecrafts. Then, the attacker selects its target spacecrafts to mount FDI attack to achieve the maximum damage. In physical layer, a Nash equilibrium (NE) control strategy is proposed for MSS to quantify system performance under the effect of attacks by solving a game problem. In cyber layer, a fuzzy Stackelberg game framework is used to examine the rivalry process between the attacker and defender. The strategies of both attacker and defender are given based on the analysis of physical layer and cyber layer. Finally, a simulation example is used to test the viability of the proposed cross layer fuzzy game algorithm.

A Multi-Type Feature Fusion Network Based on Importance Weighting for Occluded Human Pose Estimation

Jiahong Jiang, Nan Xia, Siyao Zhou

2025, 12(4): 789-805. doi: 10.1109/JAS.2024.124953

Abstract(135) HTML (27) PDF(26)

Abstract:
Human pose estimation is a challenging task in computer vision. Most algorithms perform well in regular scenes, but lack good performance in occlusion scenarios. Therefore, we propose a multi-type feature fusion network based on importance weighting, which consists of three modules. In the first module, we propose a multi-resolution backbone with two feature enhancement sub-modules, which can extract features from different scales and enhance the feature expression ability. In the second module, we enhance the expressiveness of keypoint features by suppressing obstacle features and compensating for the unique and shared attributes of keypoints and topology. In the third module, we perform importance weighting on the adjacency matrix to enable it to describe the correlation among nodes, thereby improving the feature extraction ability. We conduct comparative experiments on the keypoint detection datasets of common objects in Context 2017 (COCO2017), COCO-Wholebody and CrowdPose, achieving the accuracy of 78.9%, 67.1% and 77.6%, respectively. Additionally, a series of ablation experiments are designed to show the performance of our work. Finally, we present the visualization of different scenarios to verify the effectiveness of our work.

Analysis of Students’ Positive Emotion and Smile Intensity Using Sequence-Relative Key-Frame Labeling and Deep-Asymmetric Convolutional Neural Network

Zhenzhen Luo, Xiaolu Jin, Yong Luo, Qiangqiang Zhou, Xin Luo

2025, 12(4): 806-820. doi: 10.1109/JAS.2024.125016

Abstract(126) HTML (32) PDF(29)

Abstract:
Positive emotional experiences can improve learning efficiency and cognitive ability, stimulate students’ interest in learning, and improve teacher-student relationships. However, positive emotions in the classroom are primarily identified through teachers’ observations and postclass questionnaires or interviews. The expression intensity of students, which is extremely important for fine-grained emotion analysis, is not considered. Hence, a novel method based on smile intensity estimation using sequence-relative key-frame labeling is presented. This method aims to recognize the positive emotion levels of a student in an end-to-end framework. First, the intensity label is generated robustly for each frame in the expression sequence based on the relative key frames to address the lack of annotations for smile intensity. Then, a deep-asymmetric convolutional neural network learns the expression model through dual neural networks, to enhance the stability of the network model and avoid the extreme attention region learned. Further, dual neural networks and the dual attention mechanism are integrated using the intensity label based on the relative key frames as the supervised information. Thus, diverse features are effectively extracted and subtle appearance differences between different smiles are perceived based on different perspectives. Finally, comparative experiments for the convergence speed, model-training parameters, confusion matrix, and classification probability are performed. The proposed method was applied to a real classroom scene to analyze the emotions of students. Numerous experiments validated that the proposed method is promising for analyzing the differences in the positive emotion of students while learning in a classroom.