IEEE/CAA Journal of Automatica Sinica

Parallel Theaters in CPSS: From Shadows of ISDOS to Intelligence of Decision Theaters

Qinghua Ni, Buday Viktória, Fei Lin, Jun Huang, Levente Kovács, Nan Zheng, Fei-Yue Wang

2025, 12(6): 1059-1062. doi: 10.1109/JAS.2025.125567

Abstract(30) HTML (7) PDF(0)

Abstract:

Switching in Sliding Mode Control: A Spatio-Temporal Perspective

Xinghuo Yu

2025, 12(6): 1063-1071. doi: 10.1109/JAS.2025.125423

Abstract(142) HTML (10) PDF(43)

Abstract:
Sliding mode control (SMC) is a widely adopted control technology known for its robustness and simplicity. The essence of SMC is to use discontinuous control to drive a system into a pre-defined motion, called the sliding mode, which is designed with desirable dynamical properties. In the sliding mode, the controlled system is insensitive to the matched uncertainties and disturbances.Most SMC theory and methods have been developed based on the dynamical systems in the continuous-time domain, where switching functions play a critical role. Ideal switching is supposed to be instantaneous, activating as soon as the switching condition is met. However, in practice, switching mechanisms are affected by imperfections such as time delays, unmodeled dynamics, defects, digitization effects, and actuation limitations, which can degrade the salient properties of SMC. Understanding these effects and developing mitigation strategies are essential for industrial applications. Furthermore, the advent of networked control environments presents new challenges like limited communication bandwidth, latency and cyberattack, which have seen the emergence of the event-triggered SMC recently. Despite these significant advances, there is a lack of comprehensive studies which examine the commonalities and distinctions of utilizing switching in SMC across the continuous-time and discrete-time domains and beyond.This paper investigates the role of switching in SMC from a spatio-temporal perspective, considering both state-space and time aspects. The aim is to facilitate better understanding of its effects and misbehaviors, and to unlock its full potential for future applications. The interplay between SMC methods in the continuous-time and discrete-time domains is analyzed, and their shared principles and unique challenges are identified. Furthermore, important technical issues relating to switching across these time domains are explored, and several myths and pitfalls in their theory and applications are depicted. The relationships of SMC with other switching-based control systems such as switched control systems, fuzzy control systems, and event-triggered control systems are discussed. The impact of networked control environments on SMC in the continuous-time and discrete-time domains is also examined. Finally, key challenges and opportunities are outlined for future work in SMC and beyond.

Innovations and Refinements in LiDAR Odometry and Mapping: A Comprehensive Review

Guangjie Liu, Kai Huang, Xiaolan Lv, Yuanhao Sun, Hailong Li, Xiaohui Lei, Quanchun Yuan, Lei Shu

2025, 12(6): 1072-1094. doi: 10.1109/JAS.2025.125198

Abstract(323) HTML (7) PDF(63)

Abstract:
Since its introduction in 2014, the LiDAR odometry and mapping (LOAM) algorithm has become a cornerstone in the fields of autonomous driving and intelligent robotics. LOAM provides robust support for autonomous navigation in complex dynamic environments through precise localization and environmental mapping. This paper offers a comprehensive review of the innovations and optimizations made to the LOAM algorithm, covering advancements in multi-sensor fusion technology, frontend processing optimization, backend optimization, and loop closure detection. These improvements have significantly enhanced LOAM’s performance in various scenarios, including urban, agricultural, and underground environments. However, challenges remain in areas such as data synchronization, real-time processing, computational complexity, and environmental adaptability. Looking ahead, future developments are expected to focus on creating more efficient multi-sensor fusion algorithms, expanding application domains, and building more robust systems, thereby driving continued progress in autonomous driving, intelligent robotics, and autonomous unmanned systems.

Embodied Multi-Agent Systems: A Review

Zhuo Li, Weiran Wu, Yunlong Guo, Jian Sun, Qing-Long Han

2025, 12(6): 1095-1116. doi: 10.1109/JAS.2025.125552

Abstract(57) HTML (9) PDF(15)

Abstract:
Multi-agent systems (MASs) have demonstrated significant achievements in a wide range of tasks, leveraging their capacity for coordination and adaptation within complex environments. Moreover, the enhancement of their intelligent functionalities is crucial for tackling increasingly challenging tasks. This goal resonates with a paradigm shift within the artificial intelligence (AI) community, from “internet AI” to “embodied AI”, and the MASs with embodied AI are referred to as embodied multi-agent systems (EMASs). An EMAS has the potential to acquire generalized competencies through interactions with environments, enabling it to effectively address a variety of tasks and thereby make a substantial contribution to the quest for artificial general intelligence. Despite the burgeoning interest in this domain, a comprehensive review of EMAS has been lacking. This paper offers analysis and synthesis for EMASs from a control perspective, conceptualizing each embodied agent as an entity equipped with a “brain” for decision and a “body” for environmental interaction. System designs are classified into open-loop, closed-loop, and double-loop categories, and EMAS implementations are discussed. Additionally, the current applications and challenges faced by EMASs are summarized and potential avenues for future research in this field are provided.

Instance by Instance: An Iterative Framework for Multi-Instance 3D Registration

Jiaqi Yang, Xinyue Cao, Xiyu Zhang, Yuxin Cheng, Zhaoshuai Qi, Siwen Quan

2025, 12(6): 1117-1128. doi: 10.1109/JAS.2024.125058

Abstract(90) HTML (6) PDF(5)

Abstract:
Multi-instance registration is a challenging problem in computer vision and robotics, where multiple instances of an object need to be registered in a standard coordinate system. Pioneers followed a non-extensible one-shot framework, which prioritizes the registration of simple and isolated instances, often struggling to accurately register challenging or occluded instances. To address these challenges, we propose the first iterative framework for multi-instance 3D registration (MI-3DReg) in this work, termed instance-by-instance (IBI). It successively registers instances while systematically reducing outliers, starting from the easiest and progressing to more challenging ones. This enhances the likelihood of effectively registering instances that may have been initially overlooked, allowing for successful registration in subsequent iterations. Under the IBI framework, we further propose a sparse-to-dense correspondence-based multi-instance registration method (IBI-S2DC) to enhance the robustness of MI-3DReg. Experiments on both synthetic and real datasets have demonstrated the effectiveness of IBI and suggested the new state-of-the-art performance with IBI-S2DC, e.g., our mean registration F1 score is 12.02%/12.35% higher than the existing state-of-the-art on the synthetic/real datasets. The source codes are available online at https://github.com/caoxy01/IBI.

Multi-Phase Degradation Modeling Based on Uncertain Random Process for Remaining Useful Life Prediction Under Triple Uncertainties

Xuerui Cao, Kaixiang Peng, Ruihua Jiao

2025, 12(6): 1129-1143. doi: 10.1109/JAS.2024.124791

Abstract(120) HTML (6) PDF(19)

Abstract:
Due to abrupt changes in the intrinsic degradation mechanism or shock from external environmental pressure, degradations of some equipment are characterized by multi-phase and jumps. Meanwhile, equipment is subject to inherent fluctuations, limited data and imperfect measurements resulting in aleatory, epistemic and measurement uncertainties of the degradation process. This paper proposes a degradation model and remaining useful life (RUL) prediction method under triple uncertainties for a category of complex equipment with multi-phase degradation and jumps. First, a multi-phase degradation model with random jumps and measurement errors is constructed based on uncertain random processes. Afterward, the analytic expression of RUL prediction considering the heterogeneity is derived by modeling the uncertainty of degradation states at change points under the concept of first hitting time. A stochastic uncertain approach is utilized for the proposed multi-phase degradation model to identify model parameters based on historical data. Furthermore, the implied degradation features are adaptively updated in online stage using similarity-based weighted stochastic uncertain maximum likelihood estimation and Kalman filtering. Finally, the effectiveness of the method is verified by simulation example and practical case.

Data-Driven Bipartite Consensus Control for Large Workpieces Rotation of Nonlinear Multi-Robot Systems

Haoran Tan, Xueming Zhang, Yaonan Wang, You Wu, Yun Feng, Zhongsheng Hou

2025, 12(6): 1144-1158. doi: 10.1109/JAS.2024.124938

Abstract(63) HTML (8) PDF(13)

Abstract:
In this paper, a novel data-driven bipartite consensus control scheme is proposed for the rotation problem of large workpieces with multi-robot systems (MRSs) under a directed communication topology. The rotation of a large workpiece is described as the MRSs with cooperation and antagonism interaction. By the signed graph theory, it is further transformed into a bipartite consensus control problem, where all followers are uniformly degenerated into the general nonlinear systems based on the lateral error model. To augment the flexibility of control protocol and improve control performance, a higher-dimensional full form dynamic linearization (FFDL) technique is committed to the MRSs. The control input criterion function consists of the data model based on FFDL and the bipartite consensus error based on the signed graph theory, and the proposed control protocol is given by optimizing this criterion function. In this way, this scheme has a higher degree of freedom and better adaptive adjustment capability while not excessively increasing the control method complexity, and it can also be compatible with other forms of dynamic linearization techniques in MRSs. Further, three matrix norm lemmas are introduced to deal with the challenges of stability analysis caused by higher matrix dimensions and more robots. Finally, the effectiveness of the proposed method is verified by numerical simulations.

Joint Super-Resolution and Nonuniformity Correction Model for Infrared Light Field Images Based on Frequency Correlation Learning

You Du, Yong Ma, Jun Huang, Xiaoguang Mei, Jinhui Qin, Fan Fan

2025, 12(6): 1159-1175. doi: 10.1109/JAS.2024.124881

Abstract(224) HTML (10) PDF(32)

Abstract:
Super-resolution (SR) for the camera array-based infrared light field (IRLF) images aims to reconstruct high-resolution sub-aperture images (SAIs) from their low-resolution counterparts. Existing SR methods mainly focus on exploiting the spatial and angular information of SAIs and have achieved promising results in the visible band. However, they fail to adaptively correct the nonuniform noise in IRLF images, resulting in over-smoothness or artifacts in their results. This study proposes a novel method that reconstructs high-resolution IRLF images while correcting the nonuniformity. The main idea is to decompose the structure and nonuniform noise into high- and low-frequency components and then learn the frequency correlations to help correct the nonuniformity. To learn the frequency correlation, intra- and inter-frequency units are designed. The former learns the correlation of neighboring pixels within each component, aiming to reconstruct the structure and coarsely remove nonuniform noise. The latter models the correlation of contents between different components to reconstruct fine-grained structures and reduce residual noise. Both units are equipped with our designed triple-attention mechanism, which can jointly exploit spatial, angular, and frequency information. Moreover, we collected two real-world IRLF-image datasets with significant nonuniformity, which can be used as a common base in the field. Qualitative and quantitative comparisons demonstrate that our method outperforms state-of-the-art approaches with a clearer structure and fewer artifacts. The code is available at https://github.com/DuYou2023/IRLF-FSR.

Knowledge Classification-Assisted Evolutionary Multitasking for Two-Task Multiobjective Optimization Problems

Xiaoling Wang, Qi Kang, MengChu Zhou, Qi Deng, Zheng Fan, Haoyue Liu

2025, 12(6): 1176-1193. doi: 10.1109/JAS.2024.125070

Abstract(55) HTML (8) PDF(9)

Abstract:
To realize Industry 5.0, manufacturers face various optimization problems that seldom appear in isolation. Evolutionary MultiTasking (EMT) is an effective method to solve multiple related problems by extracting and utilizing common knowledge. Knowledge transfer is the key to the effectiveness of EMT. Existing EMT methods mainly focus on designing effective inter-task learning methods and ignore the fact that provided knowledge’s appropriateness also has a significant effect on EMT’s performance. There is plentiful knowledge in assistant tasks, and knowledge transfer may not work well and even lead to a negative effect if useless knowledge is selected to guide target tasks. EMT is thus confronted with a challenge to find appropriate knowledge. This work proposes an efficient knowledge classification-assisted EMT framework to identify and select valuable knowledge from assistant tasks. During the evolution process, better-performing candidates are supposed to have advantages in exploitation. Therefore, assistant individuals that are similar to better-performing target individuals are used to provide positive knowledge. Specifically, the target sub-population is divided into different levels and then a classifier is trained to divide assistant sub-population. Considering that target and assistant sub-populations have different characteristics, we use domain adaptation to reduce their distribution discrepancies. In this way, the trained classifier can classify assistant individuals more accurately, and truly useful knowledge can be selected for target tasks. The superior performance of our proposed framework over state-of-the-art algorithms is verified via a series of benchmark problems.

Self-Cumulative Contrastive Graph Clustering

Xiaoqiang Yan, Kun Deng, Quan Zou, Zhen Tian, Hui Yu

2025, 12(6): 1194-1208. doi: 10.1109/JAS.2024.125025

Abstract(178) HTML (7) PDF(33)

Abstract:
Contrastive graph clustering (CGC) has become a prominent method for self-supervised representation learning by contrasting augmented graph data pairs. However, the performance of CGC methods critically depends on the choice of data augmentation, which usually limits the capacity of network generalization. Besides, most existing methods characterize positive and negative samples based on the nodes themselves, ignoring the influence of neighbors with different hop numbers on the node. In this study, a novel self-cumulative contrastive graph clustering (SC-CGC) method is devised, which is capable of dynamically adjusting the influence of neighbors with different hops. Our intuition is that better neighbors are closer and distant ones are further away in their feature space, thus we can perform neighbor contrasting without data augmentation. To be specific, SC-CGC relies on two neural networks, i.e., autoencoder network (AE) and graph autoencoder network (GAE), to encode the node information and graph structure, respectively. To make these two networks interact and learn from each other, a dynamic fusion mechanism is devised to transfer the knowledge learned by AE to the corresponding GAE layer by layer. Then, a self-cumulative contrastive loss function is designed to characterize the structural information by dynamically accumulating the influence of the nodes with different hops. Finally, our approach simultaneously refines the representation learning and clustering assignments in a self-supervised manner. Extensive experiments on 8 realistic datasets demonstrate that SC-CGC consistently performs better over SOTA techniques. The code is available at https://github.com/Xiaoqiang-Yan/JAS-SCCGC.

Non-Singular Practical Fixed-time Prescribed Performance Adaptive Fuzzy Consensus Control for Multi-Agent Systems Based on an Observer

Chi Ma, Dianbiao Dong

2025, 12(6): 1209-1220. doi: 10.1109/JAS.2024.124428

Abstract(201) HTML (12) PDF(62)

Abstract:
In this paper, the problem of non-singular fixed-time control with prescribed performance is studied for multi-agent systems characterized by uncertain states, nonlinearities, and non-strict feedback. To mitigate the nonlinearity, a fuzzy logic algorithm is applied to approximate the intrinsic dynamics of the system. Furthermore, a fuzzy logic system state observer based on leader state information is designed to address the partial unobservability of followers. Subsequently, the power integral method is incorporated into the backstepping approach to avoid singularities in the fixed-time controller. A command filter method is introduced into the standard backstepping approach to reduce the computational complexity of controller design. Then, a non-singular fixed-time adaptive control strategy with prescribed performance is proposed by constraining the tracking error within a prescribed range. Rigorous theoretical analysis ensures the convergence of consensus error in the multi-agent system to the prescribed performance region within a fixed time. Finally, the practicality of the algorithm is validated through numerical simulations.

SILIC: Intelligent On/Off Control for Networked Solar Insecticidal Lamps

Heyang Yao, Lei Shu, Yuli Yang, Miguel Martínez-García, Wei Lin

2025, 12(6): 1221-1235. doi: 10.1109/JAS.2024.124668

Abstract(213) HTML (6) PDF(29)

Abstract:
The solar insecticidal lamp (SIL) is an innovative green control device. Nevertheless, a major challenge is often encountered when carrying out insecticidal work is low energy utilization efficiency. The substantial energy consumption required to turn on the SIL, coupled with the extension of insecticidal working time during the low pest activity periods, can result in low energy efficiency. Especially when the energy storage level is below 50%, the inefficient use of energy significantly reduces the effectiveness of pest control. Consequently, an ineffective on/off scheme for these lamps may lead to suboptimal energy utilization. In this paper, we present the solar insecticidal lamp intelligent energy management scheme (SIL-IEMS) to address the challenge of inefficient energy utilization in the solar insecticidal lamp internet of things (SIL-IoT). SIL-IEMS primarily utilizes genetic algorithm (GA) and greedy algorithms to optimize insecticidal working time by considering constraints such as residual energy and the number of trap pests. Comparing SIL-IEMS to the traditional remote switching method (TRSM) and the solar insecticidal lamp genetic algorithm (SILGA), our simulation results showcase its superior energy efficiency and pest control effectiveness. Particularly noteworthy is the SILIEMS’s 17.6% increase in insecticidal efficiency compared to TRSM and 6% improvement over SILGA when the SIL begins with a remaining energy level of 15%.

Optimization Algorithms Based on Double-Integral Coevolutionary Neurodynamics in Deep Learning

Dan Su, Jie Han, Chunhua Yang, Weihua Gui

2025, 12(6): 1236-1245. doi: 10.1109/JAS.2025.125210

Abstract(109) HTML (8) PDF(25)

Abstract:
Deep neural networks are increasingly exposed to attack threats, and at the same time, the need for privacy protection is growing. As a result, the challenge of developing neural networks that are both robust and capable of strong generalization while maintaining privacy becomes pressing. Training neural networks under privacy constraints is one way to minimize privacy leakage, and one way to do this is to add noise to the data or model. However, noise may cause gradient directions to deviate from the optimal trajectory during training, leading to unstable parameter updates, slow convergence, and reduced model generalization capability. To overcome these challenges, we propose an optimization algorithm based on double-integral coevolutionary neurodynamics (DICND), designed to accelerate convergence and improve generalization in noisy conditions. Theoretical analysis proves the global convergence of the DICND algorithm and demonstrates its ability to converge to near-global minima efficiently under noisy conditions. Numerical simulations and image classification experiments further confirm the DICND algorithm’s significant advantages in enhancing generalization performance.

A Proportional Integral Controller-Enhanced Non-Negative Latent Factor Analysis Model

Ye Yuan, Siyang Lu, Xin Luo

2025, 12(6): 1246-1259. doi: 10.1109/JAS.2024.125055

Abstract(42) HTML (8) PDF(14)

Abstract:
A non-negative latent factor (NLF) model is able to be built efficiently via a single latent factor-dependent, non-negative and multiplicative update (SLF-NMU) algorithm for performing precise representation to high-dimensional and incomplete (HDI) matrix from many kinds of big-data-related applications. However, an SLF-NMU algorithm updates a latent factor relying on the current update increment only without considering past learning information, making a resultant model suffer from slow convergence. To address this issue, this study proposes a proportional integral (PI) controller-enhanced NLF (PI-NLF) model with two-fold ideas: 1) Designing an increment refinement (IR) mechanism, which formulates the current and past update increments as the proportional and integral terms of a PI controller, thereby assimilating the past update information into the learning scheme smoothly with high efficiency; 2) Deriving an IR-based SLF-NMU (ISN) algorithm, which updates a latent factor following the principle of an IR mechanism, thus significantly accelerating an NLF model’s convergence rate. The simulation results on eight HDI matrices collected by real applications validate that a PI-NLF model outstrips several leading-edge models in both computational efficiency and accuracy when estimating missing data within an HDI matrix. The proposed PI-NLF model can be effectively applied to applications involving HDI matrix like e-commerce system, social network, and cloud service system. The code is available at https://github.com/yuanyeswu/PINLF/blob/main/PINLF-code.zip.

Hazard-Aware Weighted Advantage Combination for UAV Target Tracking and Obstacle Avoidance

Lele Xu, Jian Liu, Xiaoguang Chang, Xuping Liu, Changyin Sun

2025, 12(6): 1260-1271. doi: 10.1109/JAS.2024.124920

Abstract(193) HTML (8) PDF(43)

Abstract:
In recent years, the rapid evolution of unmanned aerial vehicles (UAVs) has brought about transformative changes across various industries. However, addressing fundamental challenges in UAV technology, particularly target tracking and obstacle avoidance, remains crucial for wildlife protection, military industry security, etc. Many existing methods based on reinforcement learning to solve UAV multi-tasks need to be redesigned and retrained, and cannot be quickly and effectively extended to other scenarios. To this end, we propose a novel solution based on a hazard-aware weighted advantage combination for UAV target tracking and obstacle avoidance. First, we independently trained the UAV target tracking and obstacle avoidance using the dueling double deep Q-network reinforcement learning algorithm. Subsequently, in a multitasking scenario, we introduce the two pre-trained networks. Meanwhile, we design a weight determined by the present risk level encountered by the UAV. This weight is utilized to perform a weighted summation of the advantage values from both networks, eliminating the need for retraining to obtain the final action. We validate our approach through extensive simulation experiments in the robotics simulator known as CoppeliaSim. The results demonstrate that our method outperforms current state-of-the-art techniques, achieving superior performance in both tracking accuracy and avoidance of collisions.

Mixed Motivation Driven Social Multi-Agent Reinforcement Learning for Autonomous Driving

Long Chen, Peng Deng, Lingxi Li, Xuemin Hu

2025, 12(6): 1272-1282. doi: 10.1109/JAS.2025.125201

Abstract(61) HTML (8) PDF(5)

Abstract:
Despite great achievement has been made in autonomous driving technologies, autonomous vehicles (AVs) still exhibit limitations in intelligence and lack social coordination, which is primarily attributed to their reliance on single-agent technologies, neglecting inter-AV interactions. Current research on multi-agent autonomous driving (MAAD) predominantly focuses on either distributed individual learning or centralized cooperative learning, ignoring the mixed-motive nature of MAAD systems, where each agent is not only self-interested in reaching its own destination but also needs to coordinate with other traffic participants to enhance efficiency and safety. Inspired by the mixed motivation of human driving behavior and their learning process, we propose a novel mixed motivation driven social multi-agent reinforcement learning method for autonomous driving. In our method, a multi-agent reinforcement learning (MARL) algorithm, called Social Learning Policy Optimization (SoLPO), which takes advantage of both the individual and social learning paradigms, is proposed to empower agents to rapidly acquire self-interested policies and effectively learn socially coordinated behavior. Based on the proposed SoLPO, we further develop a mixed-motive MARL method for autonomous driving combined with a social reward integration module that can model the mixed-motive nature of MAAD systems by integrating individual and neighbor rewards into a social learning objective for improved learning speed and effectiveness. Experiments conducted on the MetaDrive simulator show that our proposed method outperforms existing state-of-the-art MARL approaches in metrics including the success rate, safety, and efficiency. Moreover, the AVs trained by our method form coordinated social norms and exhibit human-like driving behavior, demonstrating a high degree of social coordination.