Xuhui Bu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Xuhui Bu has a total of 146 co-authors. Here are the top co-authors by number of shared publications:
- Zhongsheng Hou
Qingdao University
33 shared publications Zhongsheng Hou
School of Automation, Qingdao University, Qingdao, China
22 shared publications- Qiongxia Yu
School of Electrical Engineering and Automation, Henan Polytechnic University, Jiaozuo, China
16 shared publications - Ting Lei
Zhengzhou University of Light Industry
11 shared publications - Fengchen Tian
Henan Polytechnic University
10 shared publications - Panpan Zhu
Henan University
10 shared publications
Latest Publications
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72  (4) , 8239-8251
Data-Driven Safety-Critical Flocking Control for Multi-Agent Systems With Unknown Dynamics Under Uncertain Environments
The rapid development of consumer electronics, including consumer drones, smart home robots, and Internet of Things (IoT) sensor networks, has spurred the deployment of multi-agent systems (MASs) in increasingly complex and uncertain real-world environments. A critical challenge in these applications is to ensure safe and coordinated operation when the system dynamics are unavailable and the environmental conditions are unknown a priori. This paper proposes a data-driven learning control barrier function (DDLCBF) framework to address the safety-critical flocking control problem for MASs with unknown nonlinear dynamics operating in uncertain environments. First, a distributed flocking controller is constructed using pseudo partial derivatives (PPD) that relies exclusively on input-output (I/O) data, eliminating the requirement of an explicit system model. The controller is further extended to cooperative-competitive networks on signed directed graphs, where bipartite flocking convergence is guaranteed under a structural balance condition via gauge transformation. Second, a Gaussian process (GP) regression-based learning scheme is developed to approximate stochastic control barrier functions (CBFs) from real-time sensor measurements, thereby enabling probabilistic safety guarantees without prior environmental knowledge. Third, the DDLCBF scheme is formulated by integrating the PPD-based data model into the CBF constraints, yielding a quadratic programming (QP) controller that simultaneously ensures safety and low environmental uncertainty. Rigorous theoretical analyses are provided, including convergence guarantees for the PPD estimator and probabilistic safety bounds. The effectiveness and practicality of the proposed approach are validated through numerical simulations and physical experiments on an indoor multi-unmanned aerial vehicle (UAV) platform equipped with a motion capture system.
- 1 November 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 72  (4) , 8239-8251
Data-Driven Safety-Critical Flocking Control for Multi-Agent Systems With Unknown Dynamics Under Uncertain Environments
The rapid development of consumer electronics, including consumer drones, smart home robots, and Internet of Things (IoT) sensor networks, has spurred the deployment of multi-agent systems (MASs) in increasingly complex and uncertain real-world environments. A critical challenge in these applications is to ensure safe and coordinated operation when the system dynamics are unavailable and the environmental conditions are unknown a priori. This paper proposes a data-driven learning control barrier function (DDLCBF) framework to address the safety-critical flocking control problem for MASs with unknown nonlinear dynamics operating in uncertain environments. First, a distributed flocking controller is constructed using pseudo partial derivatives (PPD) that relies exclusively on input-output (I/O) data, eliminating the requirement of an explicit system model. The controller is further extended to cooperative-competitive networks on signed directed graphs, where bipartite flocking convergence is guaranteed under a structural balance condition via gauge transformation. Second, a Gaussian process (GP) regression-based learning scheme is developed to approximate stochastic control barrier functions (CBFs) from real-time sensor measurements, thereby enabling probabilistic safety guarantees without prior environmental knowledge. Third, the DDLCBF scheme is formulated by integrating the PPD-based data model into the CBF constraints, yielding a quadratic programming (QP) controller that simultaneously ensures safety and low environmental uncertainty. Rigorous theoretical analyses are provided, including convergence guarantees for the PPD estimator and probabilistic safety bounds. The effectiveness and practicality of the proposed approach are validated through numerical simulations and physical experiments on an indoor multi-unmanned aerial vehicle (UAV) platform equipped with a motion capture system.
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/CAA Journal of Automatica Sinica
- Vol. 13  (10) , 2378-2392
Region-Focused Evolutionary Search for Dynamic Multi-Objective Optimization
- 1 October 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE/CAA Journal of Automatica Sinica
- Vol. 13  (10) , 2378-2392
Region-Focused Evolutionary Search for Dynamic Multi-Objective Optimization
- 29 September 2026
- Published by SAGE Publications in Transactions of the Institute of Measurement and Control
Dual-channel dynamic event-triggered iterative learning control for nonlinear multi-agent formation control under switching topologies
This paper addresses the problems of high computational burden and excessive communication resource consumption in formation control of a class of nonlinear multi-agent systems subject to switching communication topologies, and proposes a dual-channel dynamic event-triggered iterative learning control algorithm. First, the formation tracking error is defined based on the switching topology, and a dynamic event-triggered mechanism is developed for both the sensor–controller and controller–actuator channels. Based on this, a distributed iterative learning formation control law is presented. Then, the dynamic evolution of the tracking error is formulated, and the contraction mapping principle is employed to prove convergence of the tracking error under arbitrary switching topologies. Finally, simulation results confirm the effectiveness of the proposed algorithm under switching topologies.
- 29 September 2026
- Published by SAGE Publications in Transactions of the Institute of Measurement and Control
Dual-channel dynamic event-triggered iterative learning control for nonlinear multi-agent formation control under switching topologies
This paper addresses the problems of high computational burden and excessive communication resource consumption in formation control of a class of nonlinear multi-agent systems subject to switching communication topologies, and proposes a dual-channel dynamic event-triggered iterative learning control algorithm. First, the formation tracking error is defined based on the switching topology, and a dynamic event-triggered mechanism is developed for both the sensor–controller and controller–actuator channels. Based on this, a distributed iterative learning formation control law is presented. Then, the dynamic evolution of the tracking error is formulated, and the contraction mapping principle is employed to prove convergence of the tracking error under arbitrary switching topologies. Finally, simulation results confirm the effectiveness of the proposed algorithm under switching topologies.
- 1 July 2026
- Published by AIP Publishing in Journal of Renewable and Sustainable Energy
- Vol. 18  (4)
Economic dispatch of interconnected renewable power systems via D3QN-based differential evolution
To address the challenges of high dimensionality, nonlinearity, and complex constraints in the economic dispatch of new interconnected power systems (ED), a differential evolution algorithm driven by a dueling double deep Q-network (D3QN-DE) is proposed. In this algorithm, the D3QN serves as the intelligent core, utilizing a five-dimensional state space to monitor population and individual rankings. This multi-dimensional perception enables the agent to adaptively tune the scaling factor and crossover probability based on the population's evolutionary state, effectively balancing global exploration and local exploitation for superior global search. Furthermore, by introducing a multi-segment reward function mechanism, the algorithm provides more precise feedback during evolution, accelerating convergence and improving solution quality. Concurrently, to address challenges posed by the uncertainty from high renewable energy integration, an ED model is constructed that incorporates interconnected spinning reserve capacity across multiple areas. This model establishes an inter-area reserve mutual assistance mechanism, considers area-specific differences in generation resources, and effectively absorbs load and renewable forecast errors. Consequently, it enhances system operational flexibility while ensuring efficient renewable energy accommodation. The algorithm is then employed to solve this dispatch model. Simulation results indicate that the D3QN-DE algorithm achieves higher convergence accuracy in multi-area scheduling, facilitating the optimal allocation of cross-area resources and reducing operating costs.
- 1 July 2026
- Published by AIP Publishing in Journal of Renewable and Sustainable Energy
- Vol. 18  (4)
Economic dispatch of interconnected renewable power systems via D3QN-based differential evolution
To address the challenges of high dimensionality, nonlinearity, and complex constraints in the economic dispatch of new interconnected power systems (ED), a differential evolution algorithm driven by a dueling double deep Q-network (D3QN-DE) is proposed. In this algorithm, the D3QN serves as the intelligent core, utilizing a five-dimensional state space to monitor population and individual rankings. This multi-dimensional perception enables the agent to adaptively tune the scaling factor and crossover probability based on the population's evolutionary state, effectively balancing global exploration and local exploitation for superior global search. Furthermore, by introducing a multi-segment reward function mechanism, the algorithm provides more precise feedback during evolution, accelerating convergence and improving solution quality. Concurrently, to address challenges posed by the uncertainty from high renewable energy integration, an ED model is constructed that incorporates interconnected spinning reserve capacity across multiple areas. This model establishes an inter-area reserve mutual assistance mechanism, considers area-specific differences in generation resources, and effectively absorbs load and renewable forecast errors. Consequently, it enhances system operational flexibility while ensuring efficient renewable energy accommodation. The algorithm is then employed to solve this dispatch model. Simulation results indicate that the D3QN-DE algorithm achieves higher convergence accuracy in multi-area scheduling, facilitating the optimal allocation of cross-area resources and reducing operating costs.
- 6 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13  (11) , 23981-23993
Dual-Channel Event-Triggered Distributed Finite-Time Model-Free Adaptive Cooperative Control for MASs With Prescribed Performance
This article investigates a data-driven finite-time cooperative control problem for nonlinear multiagent systems (MASs) and develops a distributed model-free adaptive prescribed-performance control (PPC) strategy based on partial-form dynamic linearization (PFDL). First, a finite-time prescribed-performance function is introduced to transform the distributed output errors, ensuring that the prescribed-performance constraints are satisfied. Next, an improved dynamic linearization method is derived to convert the transformed error dynamics into an equivalent linear data-driven model, which facilitates controller synthesis. On this basis, a finite-time performance index is constructed, and a distributed model-free adaptive cooperative control algorithm is proposed. Moreover, a dual-channel event-triggering mechanism with separate triggering rules for control inputs and measured outputs is designed to reduce the communication burden. Rigorous analysis establishes finite-time convergence of the distributed output errors and boundedness of the control inputs. Numerical simulations demonstrate the effectiveness and robustness of the proposed method.
- 6 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13  (11) , 23981-23993
Dual-Channel Event-Triggered Distributed Finite-Time Model-Free Adaptive Cooperative Control for MASs With Prescribed Performance
This article investigates a data-driven finite-time cooperative control problem for nonlinear multiagent systems (MASs) and develops a distributed model-free adaptive prescribed-performance control (PPC) strategy based on partial-form dynamic linearization (PFDL). First, a finite-time prescribed-performance function is introduced to transform the distributed output errors, ensuring that the prescribed-performance constraints are satisfied. Next, an improved dynamic linearization method is derived to convert the transformed error dynamics into an equivalent linear data-driven model, which facilitates controller synthesis. On this basis, a finite-time performance index is constructed, and a distributed model-free adaptive cooperative control algorithm is proposed. Moreover, a dual-channel event-triggering mechanism with separate triggering rules for control inputs and measured outputs is designed to reduce the communication burden. Rigorous analysis establishes finite-time convergence of the distributed output errors and boundedness of the control inputs. Numerical simulations demonstrate the effectiveness and robustness of the proposed method.
- 17 February 2026
- Published by Elsevier in Journal of the Franklin Institute
- Vol. 363  (6)
Dynamic event-triggered formation-containment of heterogeneous multi-agent systems under DoS attacks via model-free adaptive control
- 17 February 2026
- Published by Elsevier in Journal of the Franklin Institute
- Vol. 363  (6)
Dynamic event-triggered formation-containment of heterogeneous multi-agent systems under DoS attacks via model-free adaptive control
- 29 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13  (8) , 16364-16374
Fully Distributed Data-Driven Consensus Tracking for Multiple High-Speed Trains With Sensor Resolution Under Round-Robin Protocol
This article introduces a fully distributed model-free adaptive control (MFAC) method, addressing the tracking problem for multiple-high-speed trains (MHSTs) featuring sensor resolution (SR) and utilizing a round-robin protocol (RRP). SR is one of the basic characteristics of almost all sensors and an important indicator of sensor performance in practical applications, but it has not been given enough attention in the study of MHSTs. Since MHSTs have unknown dynamical properties, first, the dynamic linearization technique is used to obtain the data relation description of MHSTs. Second, to address the issue of data transmission pressure caused by limited network bandwidth in train-to-train (T2T) direct communication, the RRP can be adopted for effective alleviation. On this basis, a distributed MFAC strategy is proposed to achieve the bounded formation tracking of MHSTs with SR under the RRP. Finally, the effectiveness of the proposed control scheme is verified by simulation examples.
- 29 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13  (8) , 16364-16374
Fully Distributed Data-Driven Consensus Tracking for Multiple High-Speed Trains With Sensor Resolution Under Round-Robin Protocol
This article introduces a fully distributed model-free adaptive control (MFAC) method, addressing the tracking problem for multiple-high-speed trains (MHSTs) featuring sensor resolution (SR) and utilizing a round-robin protocol (RRP). SR is one of the basic characteristics of almost all sensors and an important indicator of sensor performance in practical applications, but it has not been given enough attention in the study of MHSTs. Since MHSTs have unknown dynamical properties, first, the dynamic linearization technique is used to obtain the data relation description of MHSTs. Second, to address the issue of data transmission pressure caused by limited network bandwidth in train-to-train (T2T) direct communication, the RRP can be adopted for effective alleviation. On this basis, a distributed MFAC strategy is proposed to achieve the bounded formation tracking of MHSTs with SR under the RRP. Finally, the effectiveness of the proposed control scheme is verified by simulation examples.
- 26 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Forensics and Security
- Vol. 21  (15566013) , 1737-1750
Event-Triggered Model-Free Adaptive Predictive Control for Networked Wind-Power Microgrids Subject to Aperiodic DoS Attacks
Escalating cybersecurity risks in communication networks pose severe challenges to the stability and reliability of microgrid control systems. This paper investigates the load frequency control problem of nonlinear networked wind-power microgrids under denial-of-service (DoS) attacks and event-triggered communication. A data-driven prediction framework is first established to construct a control-oriented dynamic linearization data model for the underlying nonlinear system. Sub-sequently, a novel event-triggered model-free adaptive predictive control (ET-MFAPC) algorithm with time-varying parameters is proposed, which integrates multi-step adaptive prediction and receding-horizon optimization. The proposed algorithm features several key innovations: 1) An output-related auxiliary variable is designed to circumvent unavailable pseudo-partial derivatives (PPDs) and their sign constraints. 2) An adaptive predictive compensation module is developed to mitigate the impacts of DoS attacks by reconstructing hijacked data packets. 3) An attack-aware online optimization mechanism is formulated to adaptively tune controller gains for improved system performance. Furthermore, a comprehensive security analysis is conducted based on a model-independent data energy function, and explicit boundary conditions are derived for tolerable attack frequency and duration. Finally, the effectiveness and robustness of the proposed method are validated through simulation studies.
- 26 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Forensics and Security
- Vol. 21  (15566013) , 1737-1750
Event-Triggered Model-Free Adaptive Predictive Control for Networked Wind-Power Microgrids Subject to Aperiodic DoS Attacks
Escalating cybersecurity risks in communication networks pose severe challenges to the stability and reliability of microgrid control systems. This paper investigates the load frequency control problem of nonlinear networked wind-power microgrids under denial-of-service (DoS) attacks and event-triggered communication. A data-driven prediction framework is first established to construct a control-oriented dynamic linearization data model for the underlying nonlinear system. Sub-sequently, a novel event-triggered model-free adaptive predictive control (ET-MFAPC) algorithm with time-varying parameters is proposed, which integrates multi-step adaptive prediction and receding-horizon optimization. The proposed algorithm features several key innovations: 1) An output-related auxiliary variable is designed to circumvent unavailable pseudo-partial derivatives (PPDs) and their sign constraints. 2) An adaptive predictive compensation module is developed to mitigate the impacts of DoS attacks by reconstructing hijacked data packets. 3) An attack-aware online optimization mechanism is formulated to adaptively tune controller gains for improved system performance. Furthermore, a comprehensive security analysis is conducted based on a model-independent data energy function, and explicit boundary conditions are derived for tolerable attack frequency and duration. Finally, the effectiveness and robustness of the proposed method are validated through simulation studies.