提出了一种基于排序操作的进化算子自适应的遗传算法，该算法中，每个个体按适应值大小进行排序，个体的选择、交叉、变异算子的概率根据个体排序值来自适应地确定，其中选择概率还随进化过程而调节。利用Markov链的分析法证明了该算法的全局收敛性。最后，实验结果表明该算法同传统的遗传算法相比不仅能收敛到全局最优解，而且具有较快的收敛速度。

提出了一种新颖的隐节点可调的变结构径向基函数网络，并应用进化规划最优地确定和调节变结构径向基函数网络隐层节点的数目及其核函数的中心和宽度，从而使网络具有在线学习和记忆新的目标模式的功能。并将该网络应用于被动声纳目标的识别和在线学习，实验结果表明基于进化规划的变结构径向基函数网络不仅改善了网络的泛化能力，而且能够有效地解决传统神经网络技术在被动声纳目标识别过程中在线学习会造成原有记忆遗忘的困难。

A genetic algorithm and recursive least squares (RLS) learning algorithm for a Gaussian radial basis function network is described, for modelling and predicting nonlinear time series. Better generalisation performance can be achieved than that of the usual clustering and RLS method.

特征分析及提取是目标分类识别的重要环节。首先应用非线性分析方法分析了舰船噪声的极限环现象，结果表明振动噪声作为舰船噪声的主要成份，其极限环在相空间上存在倍周期或混沌行为。其次，利用分形维数和分布密度比来描绘舰船噪声在相平面上极限环的奇异性和空间形状，并给出了一种新的分维数计算算法。最后，以此作为舰船目标的特征参数送入神经网络分类器用于分类识别水面和水下两大类目标。实验结果表明，从噪声极限环中提取的非线性特征能较准确地区分我们现有的水面和水下两大类目标。为舰船噪声信号的特征提取提供了一条新的思路。

Telerobotics, the body of science and technology which bridge lannan control and purely autonomous machi-nes, is expected to be a merging point of modern develop-ments in robotics, control theory, cognitive science and computer science. Telerobot helps human operator feel and control the temote pre-unknoum environment as if he/she is at the temote site, which is usually called" telepresence". However, the transmission time delay be-tween master manipulator and slave manipulator causes in-stability of the telerobot sysem and bad" telepresence" of human operator. In this paper, a new adaptive and pas-sive control scheme based on active impedance matching is proposed to quqrantee stability and transparency of the bi-lateral telerobot system. An adaptive impedance term which is active and should adapt to the impedance of envi-toninent is defined in this new control scheme. A genetic algorithm (GA) is used to directly ulentify parameters of the environment impedance, because the GA can suc-cess in the search for global. optinaan when the search space is not differentiable or linear in the parameters. The tesults of experiment show that.