【作者】 刘畅；

【导师】 徐雍；

【作者基本信息】 广东工业大学 ， 控制科学与工程， 2022， 博士

【摘要】 随着计算机技术、网络通信技术和控制技术的发展,非线性系统分布呈网络化趋势并被广泛部署与应用,网络化与非线性系统的结合给非线性系统的研究带来了突破性的发展和动力,相应的也为非线性系统在网络环境下的分析和研究带了新的问题和挑战。本文重点研究网络化离散非线性系统在系统受限、传输受限等影响下的有限时间状态估计与同步问题。为了处理以上问题,本文提出了马尔可夫非对称区间时变建模方法、部分测量值传输调度方案、周期内耦合策略和周期系统有限时间有界问题分析等新方法,并结合远程车辆跟踪问题,验证实际应用中有限时间方法的理论可实现性。具体研究内容概括如下:1.针对马尔可夫BAM神经网络,考虑非对称的系统模态依赖的时变时延,即时变时延的间隔是系统模态依赖的。由于传感器不可避免地受到测量环境的影响,同时又间接受到系统模态的影响,采用转移概率同时依赖于系统模态和测量环境的马尔可夫链来描述非恒定测量。设计了一个凸多面体不确定性描述的非脆弱估计器,有效提高了估计的鲁棒性,降低了保守性,进而得到了在一定条件下保证估计误差系统满足随机有限时间有界的充分条件,并给出了估计器增益设计方法。2.针对带有混合时变时延的主从周期神经网络,考虑传感器能量约束造成的网络带宽限制,采用了传输信道平均分配方法,并对每个信道使用加权误差最大优先传输协议,使得主神经网络在有限的带宽资源下同一时刻有多个节点发送数据。设计了一个依赖于加权误差最大优先传输协议的周期控制器,以保证从周期神经网络能够与主周期神经网络同步。通过构造周期Lyapunov-Krasovskii函数,建立了同步误差系统有限时间有界的充分条件。3.针对具有分布式时延的周期T-S模糊主从神经网络,设计了与主从神经网络模糊规则同源的模糊控制器,并得出同步误差系统。通过构造模糊规则依赖的Lyapunov函数和利用有限时间有界分析方法,得到了同步误差系统有限时间有界的充分条件。考虑到周期特性对系统边界的影响,设计一种新的求解系统边界的算法。进而在所求边界的基础上,得到主从神经网络同步控制器增益。通过数值算例验证了模糊控制器和边界迭代算法的有效性。4.针对时变耦合神经网络,考虑节点能量不足的问题,提出一种周期性内耦合策略以减少节点之间的信息交换。通过引入一种依赖于内耦合模态的马尔可夫驱动的传输方案,即每次只允许部分测量值被传输,有效克服了神经网络与估计器之间的传输带宽约束。此外,设计了依赖于被传输数据的时变马尔可夫估计器,并基于递归矩阵不等式方法得出时变估计误差系统满足有限域_∞性能的充分条件。最后给出了估计器增益设计方法。5.针对远程车辆轨迹跟踪问题,考虑监测传感器数量有限和不可靠的通信网络。采用饱和函数来描述车辆的有界时变加速度,引入一组矩阵来描述传感器监测情况,并定义为捕获状态。利用一个时变和转移概率部分未知的马尔可夫链来分析状态的捕获情况。基于传感器捕获的不可靠车辆信息,设计了捕获状态依赖的非脆弱估计器,并推导出估计误差系统,进而得到了估计误差系统满足有限域_∞性能的充分条件。更多还原

【Abstract】 With the development of computer technology,network communication technology and control technology,nonlinear systems are distributed in a networked trend and are widely de-ployed and applied.The combination of networked and nonlinear systems has brought break-through and dynamics to the study of nonlinear systems.Correspondingly,it brings new prob-lems and challenges to the analysis and research of nonlinear systems in the network environ-ment.This thesis focuses on the finite-time state estimation and synchronization of networked discrete nonlinear systems under the influence of system constrainted and transmission con-strainted.In order to deal with the problems above,this thesis proposes new methods such as Markov asymmetric interval time-varying modeling method,partial measurement trans-mission scheduling scheme,periodic inner-coupling strategy and finite-time bounded problem analysis method for periodic systems.Combined with remote vehicle tracking,the practical feasibility of finite-time methods is verified.The specific research contents are summarized as follows:1.For the Markov BAM neural networks,the asymmetrical system mode-dependent time-varying delays are considered,which means that the interval of time-varying delays is system mode-dependent.Because the sensors are inevitably influenced by the measurement environ-ments and indirectly influenced by the system mode,a Markov chain,whose transition prob-ability matrix is system mode-dependent,is used to describe the inconstant measurement.A non-fragile estimator described by the uncertainty of a convex polyhedron is designed,which effectively improves the robustness of the estimation and reduces the conservatism.Then suf-ficient conditions are obtained to ensure that the estimation error system satisfies the stochastic finite-time bounded under certain conditions,and the design method of estimator gain is given.2.For the master-slave periodic neural networks with mixed time-varying delays,con-sidering the network bandwidth limitation caused by the energy constraints of the sensor,the average distribution method of transmission channels is introduced,and the weighted-try-once-discard protocol is adopted for each channel,which enables multiple nodes in the master neural networks can send data at the same time under limited bandwidth resources.A periodic con-troller that relies on the weighted-try-once-discard protocol protocol is designed to ensure that the slave periodic neural networks can be synchronized with the master periodic neural net-works.By constructing the periodic Lyapunov-Krasovskii function,a sufficient condition for the finite-time bounded synchronization error system is established.3.For the periodic T-S fuzzy master-slave neural networks with distributed time delay,a fuzzy controller where the fuzzy rules are homologous to the master-slave neural networks is designed,and a synchronization error system is derived.By constructing the fuzzy rule-dependent Lyapunov function and using the finite-time bounded analysis method,sufficient conditions are established to guarantee that the synchronization error system is finite-time bounded.Considering the influence of periodic characteristics on the system boundary,a new algorithm for solving the system boundary is designed.Then,on the basis of the obtained boundary,the gain of the master-slave neural networks synchronous controller is obtained.The effectiveness of the fuzzy controller and the iterative algorithm for the boundary is veri-fied by numerical examples.4.For the time-varying coupled neural networks,considering the problem of insuffi-cient node energy,a periodic inner-coupling strategy is proposed to reduce the information exchanged among nodes.By introducing a Markov-driven transmission scheme that relies on inner-coupling modes,that is,only part of the measurements are allowed to be transmitted at a time,the transmission bandwidth constraints between the neural networks and the estimator can be overcome effectively.Then,a time-varying Markov estimator that depends on the trans-mitted data is designed.Afterward,a recursive matrix inequality-based sufficient condition is established to ensure that the time-varying estimation error system meets the finite-horizon_∞performance,and the estimator gains are designed.5.For the remote vehicle tracking issue,constrained monitoring sensors and unreliable communication networks are considered.A saturation function is used to describe the bounded time-varying acceleration of the vehicle.A set of matrices are introduced to model the sen-sor monitoring conditions called captured states,and a Markov chain with time varying and partially unknown transition probability is proposed to analyze the conditions of the captured states.Then,a captured states dependent nonfragile estimator is designed based on the mea-sured unreliable vehicle information,and the estimation error system is derived.Sufficient conditions are established to ensure that the estimation error system satisfies the finite-horizon_∞performance.更多还原