基于小波神经网络的自适应逆控制及其应用

神经网络控制特别适用于具有非线性和不确定性因素的系统。采用小波神经网络(WNN)对飞行仿真转台的直流伺服系统进行实时辨识,得到其逆模型。然后将这一训练后的网络作为前馈控制器与常规反馈控制器结合构成并行自适应逆控制器,控制转台跟踪指定的速度和位置轨线。仿真结果表明该方法的有效性。     

一种基于径向基函数网络的盲波束形成方法研究

针对阵列信号处理的特点 ,提出了一种基于径向基函数网络 (RBFNN)的盲波束形成方法。该方法把高阶累积量可消除高斯噪声干扰这一特性和神经网络可并行计算的结构优势有机地结合起来 ,利用高阶累积量估计阵列的导向矢量 ,引入径向基函数网络逼近波束形成的权矢量 ,实现盲波束形成。理论分析和仿真实验表明 ,提出的算法能有效地提高运行速度 ,对系统误差也具有很强的鲁棒性。     

周期信号的孔径抖动及其去除

孔径抖动是指模数转换器采样周期之间出现的相位抖动 ,是由各时钟周期边沿出现时刻的不确定性导致的。当输入信号频率较高时孔径抖动是模数转换器信噪比下降的一个重要原因。对数据采集系统中的孔径抖动进行了仿真实验 ,介绍了反卷积算法 ,并利用反卷积算法对周期信号的孔径抖动进行了去除。仿真结果表明该方法是有效可行的     

无偏灰色GM(1,1)模型的直接建模法

针对灰色GM(1,1)模型的建模方法存在偏差,模型不满足协调性条件,不具有线性变换一致性,且通过累加生成建模时,原始序列第一点信息没有起作用等问题,提出了GM(1,1)的3种无偏灰微分方程形式,给出了无偏GM(1,1)的直接建模法,证明了新方法不仅具有白指数律重合性,而且具有线性变换一致性。实例分析表明,新方法提高了建模的精度,扩大了模型的适用范围,充分利用了原始序列的第一点信息。     

ENTROPY,INFORMATION,NOISE—STUDIES ON SYSTEM EVOLUTION

Universe, ecosystem, social system, etc. are evolving systems. The evolving processes of thesesystems have gradual small changes and rapid drastic changes with uncertainties under the constraintsof environment. Systems, as a whole, are evolving toward complexity, diversity and variety withfluctuations and jumps. New order emerges from "mutations". The evolution is bothchance-dependent and path-dependent. In this paper three basics: entropy, information and noise areemphasized with regard to system evolution which is a field that can give people wisdom to solvesystem problems with domain knowledge.     

COMPARISON OF MAX—MIN APPROACH AND NN METHOD FOR RELIABILITY OPTIMIZATION OF SERIES—PARALLEL SYSTEM

Two heuristics, the max-min approach and the Nakagawa and Nakashima method, are consideredfor the redundancy allocation problem with series-parallel structure. The max-min approach canformulate the problem as an integer linear programming problem instead of an integer nonlinearproblem. This paper presents a comparison between those methods from the standpoint of solutionquality and computational complexity. The experimental results show that the max-min approach issuperior to the Nakagawa and Nakashima method in terms of solution quality in small-scale problems,but analysis of computational complexity shows that the max-min approach is inferior to other greedyheuristics.     

STABILITY CRITERIA FOR A CLASS OF UNCERTAINSYSTEMS WITH TIME—DELAY

Some stability criteria are obtained for a class of uncertain systems with time-delay usingLyapunov functional and analytic techniques. It is easy to check the criteria by making use of theboundedness of the uncertainties.     

STRATEGIC EXERCISE OF REAL OPTIONS：INVESTMENT DECISIONS IN TECHNOLOGICAL SYSTEMS

Viewing investment projects in new technologies as real options, this paper studies the effects ofendogenous competition and asymmetric information on the strategic exercise of real options. We firstdevelop a multi-period, game-theoretic model and show how competition leads to early exercise andaggressive investment behaviors and how competition erodes option values. We then relax the typicalfull-information assumption found in the literature and allow information asymmetry to exist acrossfirms. Our model shows, in contrast to the literature that payoff is independent of the ordering ofexercise, that the sequential exercise of real options may generate both informational and payoffexternalities. We also find some surpising but interesting results such as having more information isnot necessarily better.     

探索曲面结构的小生境遗传算法

提出了具有适应值曲面结构自学习能力的多区域并行局部搜索算子PLS和约束交叉算子GC，定性地分析了它们的作用机制，给出了基子仿真实验数据的遗传算法收敛速度和全局收敛可靠性的定量计算方法。仿真实验数据表明，PLS能有效地提高搜索速度并维持足够的种群基因多样度，GC可有效地微调解的质量，使改进的遗传算法的全局收敛速度和收敛可靠性均显著地优于标准遗传算法，并具有良好的鲁棒性和稳定性．     

C4ISR系统体系结构设计及辅助工具

通过深入分析美军C^4ISR体系结构的相关研究成果，详细研究体系结构设计过程中的主要工作，针对现有C^4ISR系统体系结构辅助设计工具的不足，提出建立C^4ISR系统体系结构集成设计环境。阐明了该集成设计环境应具备的主要功能，并设计了它的逻辑结构。