{"title":"基于强化学习的DoS攻击下多智能体系统规定时间人在环最优同步控制。","authors":"Zongsheng Huang,Tieshan Li,Yue Long,Hongjing Liang","doi":"10.1109/tnnls.2025.3583248","DOIUrl":null,"url":null,"abstract":"The prescribed-time (PT) human-in-the-loop (HiTL) optimal synchronization control problem for multiagent systems (MASs) under link-based denial-of-service (DoS) attacks is investigated. First, the HiTL framework enables the human operator to govern the MASs by transmitting commands to the leader. The link-based DoS attacks cause communication blockages between agents, resulting in topology switching. Under the switching communication topology, a fully distributed observer is proposed for each follower, which simultaneously integrates a prescribed finite-time function to estimate the leader's output within the PT. This observer is characterized by a bounded gain at the PT point and guarantees global practical PT convergence, while avoiding the use of global topology information. By combining the follower dynamics with the proposed observer, an augmented system is developed. Subsequently, the model-free Q-learning algorithm is used to learn the optimal synchronization policy directly from real system data. To reduce computational burden, the Q-learning algorithm is implemented using a single critic neural network (NN) structure, with the least-squares method applied to train the NN weights. The convergence of the Q-functions generated by the proposed Q-learning algorithm is proven. Finally, simulation results verify the effectiveness of the proposed control scheme.","PeriodicalId":13303,"journal":{"name":"IEEE transactions on neural networks and learning systems","volume":"41 1","pages":""},"PeriodicalIF":8.9000,"publicationDate":"2025-07-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Prescribed-Time Human-in-the-Loop Optimal Synchronization Control for Multiagent Systems Under DoS Attacks via Reinforcement Learning.\",\"authors\":\"Zongsheng Huang,Tieshan Li,Yue Long,Hongjing Liang\",\"doi\":\"10.1109/tnnls.2025.3583248\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"The prescribed-time (PT) human-in-the-loop (HiTL) optimal synchronization control problem for multiagent systems (MASs) under link-based denial-of-service (DoS) attacks is investigated. First, the HiTL framework enables the human operator to govern the MASs by transmitting commands to the leader. The link-based DoS attacks cause communication blockages between agents, resulting in topology switching. Under the switching communication topology, a fully distributed observer is proposed for each follower, which simultaneously integrates a prescribed finite-time function to estimate the leader's output within the PT. This observer is characterized by a bounded gain at the PT point and guarantees global practical PT convergence, while avoiding the use of global topology information. By combining the follower dynamics with the proposed observer, an augmented system is developed. Subsequently, the model-free Q-learning algorithm is used to learn the optimal synchronization policy directly from real system data. To reduce computational burden, the Q-learning algorithm is implemented using a single critic neural network (NN) structure, with the least-squares method applied to train the NN weights. The convergence of the Q-functions generated by the proposed Q-learning algorithm is proven. Finally, simulation results verify the effectiveness of the proposed control scheme.\",\"PeriodicalId\":13303,\"journal\":{\"name\":\"IEEE transactions on neural networks and learning systems\",\"volume\":\"41 1\",\"pages\":\"\"},\"PeriodicalIF\":8.9000,\"publicationDate\":\"2025-07-08\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"IEEE transactions on neural networks and learning systems\",\"FirstCategoryId\":\"94\",\"ListUrlMain\":\"https://doi.org/10.1109/tnnls.2025.3583248\",\"RegionNum\":1,\"RegionCategory\":\"计算机科学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"IEEE transactions on neural networks and learning systems","FirstCategoryId":"94","ListUrlMain":"https://doi.org/10.1109/tnnls.2025.3583248","RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE","Score":null,"Total":0}
Prescribed-Time Human-in-the-Loop Optimal Synchronization Control for Multiagent Systems Under DoS Attacks via Reinforcement Learning.
The prescribed-time (PT) human-in-the-loop (HiTL) optimal synchronization control problem for multiagent systems (MASs) under link-based denial-of-service (DoS) attacks is investigated. First, the HiTL framework enables the human operator to govern the MASs by transmitting commands to the leader. The link-based DoS attacks cause communication blockages between agents, resulting in topology switching. Under the switching communication topology, a fully distributed observer is proposed for each follower, which simultaneously integrates a prescribed finite-time function to estimate the leader's output within the PT. This observer is characterized by a bounded gain at the PT point and guarantees global practical PT convergence, while avoiding the use of global topology information. By combining the follower dynamics with the proposed observer, an augmented system is developed. Subsequently, the model-free Q-learning algorithm is used to learn the optimal synchronization policy directly from real system data. To reduce computational burden, the Q-learning algorithm is implemented using a single critic neural network (NN) structure, with the least-squares method applied to train the NN weights. The convergence of the Q-functions generated by the proposed Q-learning algorithm is proven. Finally, simulation results verify the effectiveness of the proposed control scheme.
期刊介绍:
The focus of IEEE Transactions on Neural Networks and Learning Systems is to present scholarly articles discussing the theory, design, and applications of neural networks as well as other learning systems. The journal primarily highlights technical and scientific research in this domain.