Question

请各位大侠们帮我翻译一部分英文好吗？

a direct (but somewhat impractical) approach appeared in Li and Feng (1987), and an indirect method is discussed in Nijmeijer and van der Schaft (1990).
Use of RNNs for dynamical plant identification and controller design has the following characteristics:
• An RNN does not require any a priori internal understanding of the plant, or any linearity assumptions. On the other hand, for successful modeling more data must be available to offset the lack of internal (first principle) system understanding.
• An RNN approximates plant dynamics through adjustment of the number of hidden nodes and the values of weights. Thus, selecting a model structure is equivalent to simply changing the number of hidden nodes. The real states of the plant are not necessary.

• Determining the values of internal weights, referred to as training of the RNN, is a particular form of nonlinear regression, for which effective distributed algorithms are available (Almeida, 1989; Pearlmutter, 1989; Williams and Zipser, 1989; You and Nikolaou, 1992). Newton-like algorithms can also be used. It should be stressed that the approximation capabilities of the RNN are limited by the number of training sets available.
The plant modeling and controller design methodology we propose in this paper is comprised of three steps:
• Model the nonlinear plant using an RNN.

• Exact-linearize the nonlinear RNN.

• Design a linear controller for the exact-linearized model, and implement it on the real plant.

Answer 1

直接(但有些不切实际的)方法出现在李、风》(1987),以及间接的方式及范Nijmeijer讨论了这个Schaft(1990)。
使用RNNs进行动力植物辨识和控制器的设计有以下特点:
•一个RNN不需要任何先验内部了解植物、或任何线性假设。另一方面,为了成功的建模更多的数据必须要能弥补缺乏内部(第一的原则,)系统的理解。
一个RNN植物动力学模拟;通过调整隐层节点的数量和价值观念的重量。因此,选择一个模型结构简单地更改相当于一些隐藏的节点。真正的状态的植物是不需要的。

•确定权重值,称为内部RNN的培训,是一种形态的非线性回归,而有效的分布式算法,Almeida供应(1989;Pearlmutter,1989;威廉姆斯和Zipser,1989,你和Nikolaou,1992)。Newton-like算法也能使用。 应该强调的是,RNN逼近能力的数量是有限的训练集可得到的。
植物建模和控制器的设计方法本文介绍由三步骤:
•模型非线性工厂用一个RNN。

•Exact-linearize非线性RNN。

•设计线性控制器模型,并实现它的exact-linearized在真正的植物。