【正文】
原則。 (Neural Network Control) 神經(jīng)網(wǎng)絡(luò)是由所謂神經(jīng)元的簡(jiǎn)單單元按并行結(jié)構(gòu)經(jīng)過可調(diào)的連接權(quán)構(gòu)成的網(wǎng)絡(luò)。神經(jīng)網(wǎng)絡(luò)的種類很多 ,控制中常用的有多層前 向 BP 網(wǎng)絡(luò) ,RBF 網(wǎng)絡(luò) ,Hopfield 網(wǎng)絡(luò)以及自適應(yīng)共振理論模型 (ART)等。 神經(jīng)網(wǎng)絡(luò)控制就是利用神經(jīng)網(wǎng)絡(luò)這種工具從機(jī)理上對(duì)人腦進(jìn)行簡(jiǎn)單結(jié)構(gòu)模擬的新型控制和辨識(shí)方法。神經(jīng)網(wǎng)絡(luò)在控制系統(tǒng)中可充當(dāng)對(duì)象的模型 ,還可充當(dāng)控制器。常見的神經(jīng)網(wǎng)絡(luò)控制結(jié)構(gòu)有 : (1)參數(shù)估計(jì)自適應(yīng)控制系統(tǒng) 。 (2)內(nèi)??刂葡到y(tǒng) 。 (3)預(yù)測(cè)控制系統(tǒng) 。 (4)模型參考自適應(yīng)系統(tǒng) 。 (5)變結(jié)構(gòu)控制系統(tǒng)。 神經(jīng)網(wǎng)絡(luò)控制的主要特點(diǎn)是 :可以描述任意非線性系統(tǒng) 。用于非線性系統(tǒng)的辨識(shí)和估計(jì) 。對(duì)于復(fù)雜不確定性問題具有自適應(yīng)能力 ??焖?優(yōu)化計(jì)算能力 。具有分布式儲(chǔ)存能力 ,可實(shí)現(xiàn)在線、離線學(xué)習(xí)。 最近有人提出以 Hopfield 網(wǎng)絡(luò)實(shí)現(xiàn)一種多分辨率體視協(xié)同算法 ,該算法以逐級(jí)融合的方式自動(dòng)完成由粗到細(xì) ,直至全分辨率的匹配和建立。又有人提出一種網(wǎng)絡(luò)自組織控制器 ,采用變斜率的最速梯度下降學(xué)習(xí)算法 ,應(yīng)用在非線性跟蹤控制中。今后需進(jìn)一步探討的問題是提高網(wǎng)絡(luò)的學(xué)習(xí)速度 ,提出新的網(wǎng)絡(luò)結(jié)構(gòu) ,創(chuàng)造出更適用于控制的專用神經(jīng)網(wǎng)絡(luò)。 外文資料翻譯 The development of modern control theory 1. The Intelligent Control (Intelligent Control) Intelligent control is the artificial intelligence and automatic control bined, is a kind of people without intervention can be independently drive intelligent machine, realize the goal of automatic control. Intelligent control attention is not on the expression of mathematical formula, calculation and processing, and on to the task and the description of the model, symbols and environment of the recognition and the knowledge base and reasoning machine design and development. Intelligent control in production process, let the puter system imitation experts or skilled operator experience, establish knowledge based generalized model, a symbol information processing, heuristic program design, knowledge representation and the learning, reasoning and decision making intelligent technology, the external environment and system process, judgment, understanding forecast and plan, make the controlled object according to certain requirements to achieve the intended purpose. Intelligent control is based on the theory of artificial intelligence, cyberics, operations research and systematics of subjects such as cross, its main features are: (1) And has knowledge of mathematical model of generalized said and mathematical model of the said in hybrid control process。 (2) The core of intelligent control in highrise control, namely the organization level, its main task is to the actual environment or processes organizations。 (3) System getting information is not only the mathematical information, more important is to text symbols, images, graphics, sounds, etc, all kinds of information. Intelligent control is in the process of development, there still exist many problems of further researches: (1) Explore a new intelligent control theory。 (2) Using voice control。 (3) Improve the system of independent learning ability and ability。 (4) Use of the existing nonlinear technical analysis of the closedloop system characteristics。 (5) Of the intelligent control system. 2. Nonlinear Control (Nonlinear Control) Nonlinear control is the theory of plex control of an important basic problem, but also a difficult subject, the development of linear system with almost parallel. The development of the nonlinear system, mathematical tools is