动态分布参数神经网络及其稳定性分析

文中将Ｈｏｐｆｉｅｌｄ型有限状态空间动态神经网络推广到无穷维状态空间动态神经网络，即动态分布参数神经网络（ＤＤＰＮＮ），并给出了它们的有界性和稳定性，尤其是，还研究了带有微分算子的多维分布参数神经网络的时空稳定性以及在保证稳定性情况下所应满足的边界条件，文中从理论上完善和发展了动态神经网络。     

A ROBUST ADAPTIVE CONTROL DESIGN FOR A CLASS OF UNCERTAIN NONLINEAR MARKOVIAN JUMP SYSTEMS

A direct robust adaptive nonlinear control scheme for a class of uncertain nonlinear Markovian jump systems with nonlinear state‐dependent uncertainty is presented. The proposed scheme is Lyapunov‐based and guarantees the global uniform ultimate boundedness of the closed‐loop system in the mean sense. A numerical example is given to demonstrate the validity of the results.     

Markov跳变系统的输出反馈鲁棒预测控制

对离散时间Markov跳变系统, 当系统状态不完全可测时, 研究了一类基于输出反馈的鲁棒模型预测控制问题. 所研究系统为准线性参数时变的, 考虑在当前时刻系统的时变参数是已知的, 将来时刻未知的情况. 综合考虑系统存在多胞不确定性和有界噪声等因素, 通过运用线性矩阵不等式方法及变量变换思想, 将无穷时域性能指标的最小最大鲁棒预测控制问题转化为具有线性矩阵不等式约束的凸优化问题, 得到了系统的输出反馈控制律. 引入二次有界概念, 在满足输入输出约束的情况下, 保证闭环系统的随机稳定性. 数值算例验证了方法的有效性.     

Iterative algorithms for discrete periodic Riccati matrix equations

In this paper, two iterative algorithms are constructed to obtain the positive definite solutions of the discrete periodic algebraic Riccati matrix equations. In these two algorithms, the estimation of the unknown matrices are updated by using the available estimation information at the current iteration step. The convergence properties of the proposed algorithms are also given. Finally, numerical examples are employed to illustrate the effectiveness of the proposed algorithms.     

An auto-pole-placement self-tuning controller†

An explicit self-tuning controller is described that incorporates closed-loop pole location as the principal design criterion for the objective function of Clarke and Gawthrop. The algorithm includes several minor extensions to improve robustness : a variable dead-time compensator to deal with unknown or time-varying delay, and a method of treating steady-state offset that avoids retaining an integrator in the loop continuously. A new technique has been developed and included in the algorithm that enables the controller to be auto-tuning in the sense that the closed-loop pole(s) and sampling period can be determined on-line so as to optimize the system set-point step response, leaving the system designer to specify only internal model orders and noise characteristics on which to initiate retuning. The resulting controller is quite robust and can easily handle unstable, non-minimum phase systems in a single-loop configuration, as is demonstrated by several simulations including a non-linear model of a fixed-bed autothermal reactor operating over such a large range that auto-tuning of the poles is a necessity.     

Higher order direct model reference adaptive control with generic uniform ultimate boundedness

This paper proposes a new higher order model reference adaptive control (HO-MRAC) approach following direct adaptive control philosophy, which estimates unknown time-varying parameters. This approach leads to a Lyapunov based conventional MRAC update law, augmented by an observer type parameter predictor dynamics. The predictor dynamics are composed of a stable known part, a feedback of the parameter error and unknown higher order parameters, which are updated using a Lyapunov based adaptive design. So, this HO-MRAC can cope with rapidly changing parameters, due to estimation of their time derivatives. Moreover, for stability analysis, a Lyapunov based generic ultimate boundedness theorem is presented, which allows for a computation of separate bounds for each state vector partition. Furthermore, this theorem formulates the explicit specification of transient and ultimate bounds, reaching time on the ultimate bounds and a set of admissible initial conditions. Two challenging illustrative examples demonstrate the effectiveness of the proposed approach.