摘 要
在實際控制系統(tǒng)中，常常會碰到一些系統(tǒng)需要大量的實驗和不斷嘗試才能達到很好的控制效果。這種控制方式帶有很大的主觀性，同時這些系統(tǒng)很難用具體的數(shù)學(xué)模型來描述。對它們的控制方法是從人的生活經(jīng)驗中得到的，而經(jīng)典控制理論在這些系統(tǒng)面前則顯得無能為力。因此如何設(shè)計行之有效的定性與定量的轉(zhuǎn)換模型就成為擺在專家學(xué)者面前的重要課題。上個世紀(jì)90年代李德毅教授成功的運用云模型推理方法，實現(xiàn)了對三級倒立擺各類平衡姿態(tài)的動態(tài)轉(zhuǎn)換，云模型理論得到了迅速發(fā)展。
云模型是用語言值表示的某個定性概念與其定量表示之間的不確定性轉(zhuǎn)換模型。它將人類用語言值表述的控制經(jīng)驗構(gòu)成規(guī)則，多條規(guī)則構(gòu)成規(guī)則庫。當(dāng)外部有輸入刺激規(guī)則時，通過云模型不確定性推理，實現(xiàn)人的不確定性推理和控制，而且，使用這種控制方式，并不需要建立控制對象的精確數(shù)學(xué)模型，僅需要的是利用定性知識推理實現(xiàn)對系統(tǒng)的有效控制。
針對云模型的概念表述、推理規(guī)則、內(nèi)在機理等問題，諸多學(xué)者進一步提出自己的研究成果。將云模型與神經(jīng)網(wǎng)絡(luò)、蟻群算法相結(jié)合，也是近幾年新的研究趨勢。將云模型引入自動控制系統(tǒng)是一種全新的嘗試。云模型用于智能控制可以較好地解決系統(tǒng)的狀態(tài)只能用不精確的定性量來描述的復(fù)雜系統(tǒng)的控制。從本質(zhì)上來說，云模型模擬的是人的不確定性智能控制方式，是一種定性控制機理。將云模型的定性推理方法引入控制系統(tǒng)中，不要求給出被控對象的精確數(shù)學(xué)模型，將人的自然語言表達的控制經(jīng)驗通過云模型定性規(guī)則控制器，可以實現(xiàn)對倒立擺的穩(wěn)定控制。近幾年來，云模型研究領(lǐng)域又取得了許多新的進展。
云模型在智能控制領(lǐng)域已經(jīng)得到了很多應(yīng)用，但這些應(yīng)用往往是在計算機上使用軟件實現(xiàn)的，計算機本身固有的缺點，體積大、成本高、功耗大等都限制了云模型在控制領(lǐng)域的進一步應(yīng)用。
首先從云模型基本概念理論出發(fā)，包括云模型的定義、正態(tài)云模型的普適性以及三熵規(guī)則等基礎(chǔ)概念，研究學(xué)習(xí)了基于規(guī)則發(fā)生器的云模型不確定性推理算法、一維云模型系統(tǒng)的通用逼近性理論和常規(guī)云模型控制器的設(shè)計方法和步驟。其次，詳細地對云模型控制理論算法在DSP上的具體實現(xiàn)的相關(guān)技術(shù)進行了研究與實現(xiàn)，主要包括通過數(shù)的定標(biāo)將浮點數(shù)據(jù)轉(zhuǎn)化為定點數(shù)據(jù)、用程序?qū)崿F(xiàn)隨機數(shù)的生成以及硬件平臺的設(shè)計等。并在自己設(shè)計的硬件平臺基礎(chǔ)上對云模型不確定性推理算法進行了實現(xiàn)與驗證。進而，針對履帶機器人巡航系統(tǒng)設(shè)計了速度云模型控制器和轉(zhuǎn)角云模型控制器，利用超聲波傳感器實時檢測障礙物距離實現(xiàn)對履帶機器人的智能行走控制。然后，在此基礎(chǔ)上研究學(xué)習(xí)了基于MATLAB開發(fā)平臺的一些模型設(shè)計技術(shù)，主要包括用C語言編寫S-Function模塊和自動代碼生成。研究了一種在MATLAB環(huán)境下通過S函數(shù)與RTW工具將云模型推理算法集成為一個專用DSP庫的方法，進而將該云模型算法庫應(yīng)用于直流電機調(diào)速系統(tǒng)和龍人寶貝車復(fù)雜環(huán)境避障系統(tǒng)。最后，對全文所做的工作進行了總結(jié)并提出了下一步的研究方向。

關(guān)鍵詞 云模型；智能控制；數(shù)字信號處理器；云模型控制器

Abstract

In practical applications, we often encounter some systems need a lot of experimentation and constantly trying to achieve good control effect. This control method with a lot of subjectivity, and these systems are difficult to use a specific mathematical model to describe. Way of control them is to get people's life experience, but the classical control theory is powerless. How to design effective conversion model (controller) of qualitative and quantitative has become an important subject which placed in front of experts and scholars. 90s of the last century, Professor Li Deyi successfully used cloud model reasoning to realize the dynamic conversion for all kinds of balance stances of three class of inverted pendulum. Cloud model theory has been developing rapidly. At present, the cloud model has made Gratifying research results in various fields.
Against the concept of cloud model expression, inference rules, inherent mechanism and other issues, many scholars further their research achievements. Cloud model will be integrated with neural networks, ant algorithms, as well as new trends of the past few years. Cloud model is a new attempt to introduce automatic control system. Cloud model for intelligent control systems can be used to solve the state can only be described by the amount of inaccurate characterization of complex systems. In essence, the cloud is the model of uncertainty intelligent control, is a qualitative control mechanism. Qualitative Reasoning of cloud model will be introduced in the control system. It is not required to give an exact mathematical model of the object. Natural language of human experience through the cloud model qualitative rules controller will be expressed and the stability of the inverted pendulum control can be achieved. In recent years, the cloud model has scored new progress in the field.
In allusion to the development condition of cloud model basic theories and intrinsic shortcomings of computers, design a cloud model controller based on digital signal processor, which can make effective control on systems come true.
First of all, with the beginning of cloud model concepts and theories, including the definition, the universality of normal cloud and thrice entropy rule and so on, research and study the uncertain reasoning algorithm based on cloud model rules generators, the approximating theory of one dimension cloud model system and the design method and steps of general cloud model controller. Second, work related technologie..