协同一致性多智能体系统的故障参数辨识与容错控制
发布时间：2019-01-03 20:20
【摘要】：协同一致性问题为分布式多智能体系统的基础问题,网络化多智能体系统可通过协同控制律实现状态一致。单个智能体的执行器故障可通过智能体间的协同行为扩散到整个系统,使整个系统失稳。然而传统的故障诊断与容错控制未考虑智能体间的协同行为,使其难于应用于协同系统。因此,本文建立一种一致性多智能体系统故障处理框架,研究该框架下的基于未知输入观测器的故障检测方法、基于极值搜索的故障参数辨识方法和基于李雅普诺夫稳定性理论的协同容错控制策略。 首先,建立包含协同项的一阶系统和二阶系统模型,并分析系统稳定性与拉普拉斯矩阵的关系。在此基础上分析系统故障、传感器故障和执行器故障对系统性能的影响,并对执行器故障进行进一步分类。针对执行器故障,通过解耦故障项矩阵,为每个智能体设计一组未知输入观测器进行故障检测,实现了快速的故障定位。分别为一阶和二阶系统的故障检测设计了对应的仿真场景,验证了所提观测器具有快速故障检测的能力。 其次,利用极值搜索不依赖于数学模型的特性,提出一种新的自适应故障参数辨识方法,即把故障参数辨识问题转化为极值搜索的参数优化问题。单故障情况,为其设计对应的代价函数,并证明了极值搜索闭环结构的参数在满足一定约束条件下,具有负梯度的收敛能力；多故障情况,基于梯度的思想设计了一种可同时辨识多个参数的极值搜索闭环结构,并分析了多种形式激励函数的参数辨识性能。针对基于梯度思想的结构收敛速度较慢的问题,提出了一种基于牛顿算法的参数辨识结构,并验证了牛顿算法具有快速的参数辨识能力。 最后,基于李雅普诺夫稳定性思想,设计调整互连权重的协同容错策略。针对单故障情况,为其设计李雅普诺夫函数,推导系统稳定性条件,并根据该稳定条件设计权重调整策略,在实现容错的基础上满足一定的性能指标,并通过水平位置一致性运动仿真验证其有效性,其中,二阶系统中隔离了故障智能体；针对多故障情况,在隔离故障智能体的基础上,设计了可重构的协同控制律,以补偿故障对系统的影响,从而实现一致性多智能体系统的故障容错。
[Abstract]:The cooperative consistency problem is the basic problem of distributed multi-agent system. The networked multi-agent system can realize state consistency through cooperative control law. The actuator fault of a single agent can spread to the whole system through cooperative behavior between agents, which makes the whole system unstable. However, the traditional fault diagnosis and fault-tolerant control do not consider the cooperative behavior between agents, which makes it difficult to be applied to collaborative systems. Therefore, in this paper, a consistent multi-agent system fault processing framework is established, and the fault detection method based on unknown input observer is studied. Fault parameter identification method based on extremum search and cooperative fault-tolerant control strategy based on Lyapunov stability theory. First, the first and second order system models are established, and the relation between the stability of the system and Laplace matrix is analyzed. On this basis, the influence of system fault, sensor fault and actuator fault on the system performance is analyzed, and the actuator fault is further classified. For actuator fault, a set of unknown input observer is designed for each agent by decoupling the fault item matrix to detect the fault, and the fast fault location is realized. The corresponding simulation scenarios are designed for the first and second order system fault detection respectively. The proposed observer has the ability of fast fault detection. Secondly, a new adaptive fault parameter identification method is proposed, in which the problem of fault parameter identification is transformed into a parameter optimization problem of extreme value search, which does not depend on the mathematical model. In the case of single fault, the corresponding cost function is designed, and it is proved that the parameters of the closed-loop structure of extremum search have the convergence ability of negative gradient under certain constraint conditions. Based on the idea of gradient, a closed-loop structure of extremum search is designed, which can identify multiple parameters at the same time, and the parameter identification performance of various forms of excitation function is analyzed. In order to solve the problem of slow convergence of structure based on gradient, a parameter identification structure based on Newton algorithm is proposed, and the fast parameter identification ability of Newton algorithm is verified. Finally, based on Lyapunov stability theory, a cooperative fault-tolerant strategy is designed to adjust interconnection weights. For a single fault, the Lyapunov function is designed, and the stability condition of the system is deduced. According to the stability condition, the weight adjustment strategy is designed, and the certain performance index is satisfied on the basis of fault tolerance. The effectiveness of the method is verified by horizontal position consistent motion simulation, in which the fault agent is isolated in the second order system. Based on isolating the fault agent, a reconfigurable cooperative control law is designed to compensate for the influence of the fault on the system, and the fault tolerance of the consistent multi-agent system is realized.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP18;TP302.8
本文编号：2399828
[Abstract]:The cooperative consistency problem is the basic problem of distributed multi-agent system. The networked multi-agent system can realize state consistency through cooperative control law. The actuator fault of a single agent can spread to the whole system through cooperative behavior between agents, which makes the whole system unstable. However, the traditional fault diagnosis and fault-tolerant control do not consider the cooperative behavior between agents, which makes it difficult to be applied to collaborative systems. Therefore, in this paper, a consistent multi-agent system fault processing framework is established, and the fault detection method based on unknown input observer is studied. Fault parameter identification method based on extremum search and cooperative fault-tolerant control strategy based on Lyapunov stability theory. First, the first and second order system models are established, and the relation between the stability of the system and Laplace matrix is analyzed. On this basis, the influence of system fault, sensor fault and actuator fault on the system performance is analyzed, and the actuator fault is further classified. For actuator fault, a set of unknown input observer is designed for each agent by decoupling the fault item matrix to detect the fault, and the fast fault location is realized. The corresponding simulation scenarios are designed for the first and second order system fault detection respectively. The proposed observer has the ability of fast fault detection. Secondly, a new adaptive fault parameter identification method is proposed, in which the problem of fault parameter identification is transformed into a parameter optimization problem of extreme value search, which does not depend on the mathematical model. In the case of single fault, the corresponding cost function is designed, and it is proved that the parameters of the closed-loop structure of extremum search have the convergence ability of negative gradient under certain constraint conditions. Based on the idea of gradient, a closed-loop structure of extremum search is designed, which can identify multiple parameters at the same time, and the parameter identification performance of various forms of excitation function is analyzed. In order to solve the problem of slow convergence of structure based on gradient, a parameter identification structure based on Newton algorithm is proposed, and the fast parameter identification ability of Newton algorithm is verified. Finally, based on Lyapunov stability theory, a cooperative fault-tolerant strategy is designed to adjust interconnection weights. For a single fault, the Lyapunov function is designed, and the stability condition of the system is deduced. According to the stability condition, the weight adjustment strategy is designed, and the certain performance index is satisfied on the basis of fault tolerance. The effectiveness of the method is verified by horizontal position consistent motion simulation, in which the fault agent is isolated in the second order system. Based on isolating the fault agent, a reconfigurable cooperative control law is designed to compensate for the influence of the fault on the system, and the fault tolerance of the consistent multi-agent system is realized.
【学位授予单位】：中南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP18;TP302.8
【参考文献】
相关期刊论文 前3条
1 张秋花;薛惠锋;吴介军;寇晓东;;多智能体系统MAS及其应用[J];计算机仿真;2007年06期
2 周东华;胡艳艳;;动态系统的故障诊断技术[J];自动化学报;2009年06期
3 刘佳;陈增强;刘忠信;;多智能体系统及其协同控制研究进展[J];智能系统学报;2010年01期
，本文编号：2399828
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2399828.html
