基于神经网络一类非线性系统自适应控制

对一类未知可控非线性仿射系统,假设其状态不完全可测。首先,利用动态回归神经网络建立其状态方程和输出方程,用Lyapunov理论证明了辩识误差∈L∞;然后,设计状态反馈控制器,使系统的输出跟踪参考输出,从理论证明了跟踪误差趋于零。最后,通过一个仿真实验验证了本方法的有效性。     

非线性离散系统的直接自适应神经网络控制器

对于一类非仿射离散时间系统,提出了一种新的自适应神经网络控制器。首先推导与原系统等价的仿射形式模型,由仿射模型推导控制律。控制律中采用一个神经网络,与传统的基于反馈线性化的自适应神经网络设计方法中采用两个神经网络相比,计算量大大减少且避免了控制器奇异问题。神经网络权值根据系统输入输出信号进行更新,另外σ项的引入,取消了为保证参数收敛持续激励的条件。系统的稳定性通过Lyapunov方法进行了分析,仿真实例验证了控制器的有效性。     

动态神经网络的非线性系统鲁棒辨识和状态观测

针对多变量仿射非线性系统辨识和状态估计问题。采用动态神经网络作为辨识器和观测器,从而将系统辨识和状态观测设计过程合二为一,给出了辨识和状态估计的系统设计方法及具体设计步骤。在考虑建模误差项影响的情况下,采用基于Lyapunov理论的权调节律确保权值估计误差和状态观测误差的一致最终有界。仿真结果验证了所提方法的有效性。     

Adaptive output-feedback control for MIMO nonlinear systems with time-varying delays using neural networks

An adaptive neural network output-feedback regulation approach is proposed for a class of multi-input-multi-output nonlinear time-varying delayed systems. Both the designed observer and controller are free from time delays. Different from the existing results, this paper need not the assumption that the upper bounding functions of time-delay terms are known, and only a neural network is employed to compensate for all the upper bounding functions of time-delay terms, so the designed controller procedure is more simplified. In addition, the resulting closed-loop system is proved to be semi-globally ultimately uniformly bounded, and the output regulation error converges to a small residual set around the origin. Two simulation examples are provided to verify the effectiveness of control scheme.     

Adaptive control of system with hysteresis using neural networks

1.INTRODUCTIONThe piezoelectric actuators are well suited for micro-position devices in precision engineering because oftheir fast response,nanometer resolution and biggerdriving force[1].However,hysteresis inherent topiezoelectric actuator severely li mits system’s perfor-mance such as giving rise to undesirable accuracy oroscillations,even leading to instability.Hysteresischaracteristics are generally nondifferentiable andusually unknown.It is a difficult task to mitigate itsharmful ef…     

Output-feedback adaptive stochastic nonlinear stabilization using neural networks

For the first time, an adaptive backstepping neural network control approach is extended to a class of stochastic nonlinear output-feedback systems. Different from the existing results, the nonlinear terms are assumed to be completely unknown and only a neural network is employed to compensate for all unknown nonlinear functions so that the controller design is more simplified. Based on stochastic LaSalle theorem, the resulted closed-loop system is proved to be globally asymptotically stable in probability....     

基于神经网络的不确定非线性时滞系统保成本控制

保成本控制问题受到了人们的关注,并取得了很多的研究成果,但这些研究均未考虑提高系统动态特性的问题,且很少涉及非线性系统。为此,将神经网络和控制理论相结合,针对一类不确定非线性时滞系统,结合系统动态特性以及非线性扰动抑制问题,提出了新的控制算法。通过将稳定度引入保成本控制中,可以在保成本控制的基础上提高系统的动态特性;利用神经网络良好的非线性逼近能力,很好地解决了系统存在任意非线性扰动时的控制问题。     

控制方向未知的输入受限非线性系统自适应模糊反步控制

针对一类输入受限控制方向未知的非线性系统,提出一种基于Lipschitz条件的自适应模糊反步控制器的设计方法。在控制器的设计过程当中,通过变换系统形式和采用Butterworth低通滤波器解决控制方向未知的问题;采用模糊系统对不确定非线性函数进行在线逼近;利用双曲正切函数和Nussbaum函数对系统输入饱和函数进行处理;将动态面法与反步法相结合解决“计算膨胀”的问题。运用Lyapunov理论分析证明设计的控制律能够使闭环系统所有信号半全局一致有界(semi-globally uniformly ultimately bounded, SGUUB)。该方法的有效性在一类通用的高超声速飞行器的攻角控制仿真中得到了验证。     

Congestion control method: improved ERICA algorithm based on available bandwidth prediction using neural networks

1.INTRODUCTION InATM(AsynchronousTransferMode)networks, theATMFromhasdefinedsixclassesofservices: constantbitrate(CBR),real timevariablebitrate (rt VBR),non real timevariablebitrate(nrt VBR),availablebitrate(ABR),unspecifiedbitrate (UBR),andguaranteedframerate(GFR)[1].A mongthese,ABRistheonlyonesevicethatcanad justitssendingrateaccordingtothefeedbackinfor mation.Manydataapplicationsarehighlyburstyand havenowayofpredictingdatatrafficrequirementsin advance,buthavewelldefined…     

Robust stability for stochastic interval delayed Hopfield neural networks

A type of stochastic interval delayed Hopfield neural networks as du(t) = [-AIu(t) WIf(t,u(t)) WIτf7τ(uτ(t)] dt σ(t, u(t), uτ(t)) dw(t) on t≥0 with initiated value u(s) = ζ(s) on - τ≤s≤0 has been studied. By using the Razumikhin theorem and Lyapunov functions, some sufficient conditions of their globally asymptotic robust stability and global exponential stability on such systems have been given. All the results obtained are generalizations of some recent ones reported in the literature for uncertain neural networks with constant delays or their certain cases.     

Adaptive tracking controller using BP neural networks for a class of nonlinear systems

An BP neural-network-based adaptive control (NNAC) design method is described whose aim is to control a class of partially unknown nonlinear systems. Making use of the online identification of BP neural networks, the results of the identification could be used into the parameters of the controller. Not only the strong robustness with respect to uncertain dynamics and nonlinearities can be obtained, but also the output tracking error between the plant output and the desired reference output can asymptotically converge to zero by Lyapunov theory in the process of this design method.And a simulation example is also presented to evaluate the effectiveness of the design.     

神经网络在质量矩导弹控制系统上的应用

以所建立的质量矩导弹数学模型为基础,通过对模型合理的简化,得到一个耦合的非线性动力学系统,由于存在参数的不确定性以及建模误差,考虑到质量矩导弹的鲁棒性要求,采用神经网络自适应控制器对系统进行补偿。通过李亚普诺夫稳定理论证明跟踪误差是指数收敛的,仿真结果验证了这种方法的有效性。     

一种非线性时变系统小波网络辨识算法

提出一种可对任意非线性时变系统进行辨识的新方法，即基于小波神经网络的带自校正移动窗的递推最小二乘算法，与现有的神经网络辨识算法不同，该算法是根据被估权值时变速度的快慢来自适应地调整移动窗的长度，以跟踪非线性时变系统的动态特性，文中推导了了算法，并将全局算法进一步推广成不含任何矩阵运算的局部算法以提高算法的实时性能，几个典型的系统辨识仿真实例显示出这种方法具有跟踪精度高和计算简便的良好性能。     

Stability for the mix-delayed Cohen-Grossberg neural networks with nonlinear impulse

<正> In this paper,the authors are concerned with the stability of the mix-delayed Cohen-Grossbergneural networks with nonlinear impulse by the nonsmooth analysis.Some novel sufficientconditions are obtained for the existence and the globally asymptotic stability of the unique equilibriumpoint,which include the well-known results on some impulsive systems and non-impulsive systems asits particular cases.The authores also analyze the globally exponential stability of the equilibriumpoint.Two examples are exploited to illustrate the feasibility and effectiveness of our results.     

多层前向神经网络带正则化因子的算法

针对权衰减递推最小二乘算法(trueweightdecayRLS,TWDRLS)每迭代一步计算复杂度和存储要求很大,基于局部线性最小二乘算法(locallinearizedleastsquaresalgorithm,LLLS)与正则化因子,给出了多层前向神经网络带正则化因子的LLLS算法,大大减小了TWDRLS算法每迭代一步计算的复杂度和存储量。实验表明,改进的算法提高了原LLLS算法的鲁棒性和泛化能力,其性能接近TWDRLS算法。     

时滞Markov跳变BAM神经网络的鲁棒稳定性

针对一类具有Markov跳变参数的时滞双向联想记忆(BAM)神经网络,研究了其在系统参数不确定情况下的鲁棒稳定性.在不要求连接权矩阵的对称性和激励函数的可微性与单调性的情况下,通过构造适当的Lyapunov泛函得到了此类神经网络均方鲁棒稳定的充分条件.该条件考虑了时滞Markov跳变神经网络中参数的不确定性,改进了现有文献中的关于时滞Markov跳变神经网络的稳定性条件,所得结果以线性矩阵不等式(LMI)的形式给出.最后,通过实例仿真验证了所得结论的有效性.     

Traffic control based on dahlin algorithm and neural network prediction in TAM networks

The propagation delay in networks has a great adverse effect on rate-based traffic control. This paper proposes the composite control based on Dab lin algorithm feedback control and neural network feedforward predictive compensation online for ABR (available bit rate) communication in ATM (asynchronous transfer mode) networks, which can overcome the adverse effect caused by the delay on the control rapidity and stability better. The theoretical analysis and simulation research show that the scheme can make sources respond to the changes of network status rapidly, avoid the congestion effectively and utilize the bandwidth sufficiently. Compared with PID (proportional-integral-derivative) control, cell loss rate is much lower, link utilization rate is much higher, and required buffer capacity is much smaller.     

基于未知输入观测器方法的非线性切换系统故障检测

基于非线性未知输入观测器方法对一类离散时间非线性切换系统的故障检测问题进行了研究。给出了两种不同结构的非线性未知输入观测器。针对每一种观测器都考虑了两种切换信号:任意时间切换信号和满足平均驻留时间的切换信号。在任意切换信号下,使用了切换Lyapunov函数的方法来设计非线性未知输入观测器的参数;在满足平均驻留时间切换信号下,使用了多Lyapunov函数方法进行观测器参数的设计。然后利用已知的非线性未知输入观测器,得到残差生成器,进而实现故障检测。通过数值仿真例子,验证了故障检测方法的有效性。     

Novel Newton''s learning algorithm of neural networks

Newton's learning algorithm of NN is presented and realized. In theory, the convergence rate of learning algorithm of NN based on Newton's method must be faster than BP's and other learning algorithms, because the gradient method is linearly convergent while Newton's method has second order convergence rate. The fast computing algorithm of Hesse matrix of the cost function of NN is proposed and it is the theory basis of the improvement of Newton's learning algorithm. Simulation results show that the convergence rate of Newton's learning algorithm is high and apparently faster than the traditional BP method's, and the robustness of Newton's learning algorithm is also better than BP method's.     

基于分布式并行计算的神经网络算法

为了提高计算性能（速度与可扩展性）,提出了一种新颖的神经网络的并行计算体系结构和计算网络权函数的训练算法。权函数是广义Chebyshev多项式和线性函数的复合函数,只需要通过代数计算就可以求得,不需要梯度下降计算或者矩阵计算。各个权函数能够独立求解,可以通过并行系统采用并行算法计算。算法可以求得全局最优点,得到反映网络误差的一个有用的表达式。此外,算法在不超过权函数总数的范围内,还具有维持加速比与并行系统中提供的处理器的数量成线性增长的能力。仿真实验结果表明,本文算法的计算性能远远优于传统算法。