第三階段是七十年代的大系統理論時期，著重解決生物系統、社會系統這樣一些眾多變量的大系統的綜合自動化問題；方法是時域法為主；重點是大系統多級遞階控制；核心裝置是網絡化的電子計算機。 從控制論的觀點看，人是巧妙，靈活的控制系統。它善于根據條件的變化而作出正確的處理。如何將人的智能應用于實際的自動控制系統中，這是個有重要意義的問題。七十年代開始，人們不僅解決社會、經濟、管理、生態環境等系統問題，而且為解決模擬人腦功能，形成了新的學科----人工智能科學，這是控制論的發展前沿。計算機技術的發展為人工智能的發展提供了堅實的基礎。人們通過計算機的強大的信息處理能力來開發人工智能，并用它來模仿人腦。在沒有人的干預下，人工智能系統能夠進行自我調節、自我學習和自我組織，以適應外界環境的變化，并作出相應的決策和控制。 科學在發展，控制論也在不斷發展。所以“現代”兩個字加在“控制理論”前面，其含義會給人誤解的。實際上，我們講的現代控制理論指的是五六十年代所產生的一些控制理論，主要包括： 用狀態空間法對多輸入多輸出復雜系統建模，并進一步通過狀態方程求解分析，研究系統的可控性、可觀性及其穩定性，分析系統的實現問題； 用變分法、大（?。┲翟?、動態規劃原理等求解系統的優控制問題；其中常見的優控制包括時間短、能耗少等等，以及它們的組合優化問題；相應的有狀態調節器、輸出調節器、跟蹤器等綜合設計問題； 優控制往往要求系統的狀態反饋控制，但在許多情況下系統的狀態是很難求得的，往往需要一些專門的處理方法，如卡爾曼濾波技術來求得。這些都是現代控制理論的范疇。 六十年代以來，現代控制理論各方面有了很大的發展，而且形成幾個重要的分支課程，如線性系統理論，優控制理論，自適應控制理論，系統辯識理論，等等。 對控制系統一定要進行定量分析，否則就沒有控制論；而要進行定量分析，就必須用數學模型來刻劃描述系統，也即建立系統的數學模型，這是一個很重要的問題。 經典控制論中常用一個高階微分方程來描述系統的運動規律，而現代控制論中采用的是狀態空間法，就是用一組狀態變量的一階微分方程組作為系統的數學模型。這是現代控制理論與經典控制理論的一個重要區別。從某種意義上說，經典控制中的微分方程只能描述系統的輸入與輸出的關系，卻不能描述系統內部的結構及其狀態變量，它描述的只是一個‘黑箱’系統。而現代控制論中的狀態空間法不但能描述系統輸入與輸出的關系，而且還能完全描述內部的結構及其狀態變量的關系，它描述的是一個‘白箱’系統。由于能夠描述更多的系統信息，所以可以實現更好的系統控制。 控制論、信息論、系統論作為獨立的學科，各自都有自己的發展方向，同時又有內在的聯系。在研究通訊和控制時，都離不開系統；研究系統或控制時，又離不開信息。一般系統論把其研究對象作為一個整體加以考慮，提出適合于一切系統的模式、原則和規律，強調系統于個體，這有助于說明有組織的系統。
The third stage is the period of large-scale system theory in the 1970s, which focuses on the comprehensive automation of large-scale systems with many variables such as biological systems and social systems; Time domain method is the main method; The emphasis is on multi-level hierarchical control of large-scale systems; The core device is a networked computer. From the viewpoint of cybernetics, man is the most ingenious and flexible control system. It is good at making correct treatment according to the change of conditions. How to apply human intelligence to practical automatic control system is a significant problem. Since the 1970s, people have not only solved the system problems of society, economy, management and ecological environment, but also formed a new discipline - Artificial Intelligence Science, which is the development frontier of cybernetics. The development of computer technology provides a solid foundation for the development of artificial intelligence. People develop artificial intelligence through the powerful information processing ability of computer, and use it to imitate the human brain. Without human intervention, the artificial intelligence system can conduct self-regulation, self-learning and self-organization to adapt to the changes of the external environment and make corresponding decisions and controls. Science is developing, and cybernetics is also developing. Therefore, if the word "modern" is added in front of "control theory", its meaning will be misunderstood. In fact, the modern control theory we are talking about refers to some control theories produced in the 1950s and 1960s, mainly including: modeling the multi input and multi output complex system with the state space method, further studying the controllability, observability and stability of the system through the solution and analysis of the state equation, and analyzing the implementation of the system; The optimal control problem of the system is solved by variational method, maximum (minimum) value principle and dynamic programming principle; The common optimal control includes the shortest time, the least energy consumption and so on, as well as their combination optimization problems; The corresponding integrated design problems include state regulator, output regulator and tracker; The optimal control often requires the state feedback control of the system, but in many cases, the state of the system is difficult to obtain, and it often needs some special processing methods, such as Kalman filter technology. These are the categories of modern control theory. Since the 1960s, various aspects of modern control theory have developed greatly, and several important branch courses have been formed, such as linear system theory, optimal control theory, adaptive control theory, system identification theory, and so on. Quantitative analysis must be carried out for the control system, otherwise there will be no cybernetics; For quantitative analysis, it is necessary to describe the system with mathematical model, that is, to establish the mathematical model of the system, which is a very important problem. In classical cybernetics, a higher-order differential equation is often used to describe the motion law of the system, while in modern cybernetics, the state space method is used, that is, a set of first-order differential equations of state variables are used as the mathematical model of the system. This is an important difference between modern control theory and classical control theory. In a sense, the differential equation in classical control can only describe the relationship between the input and output of the system, but can not describe the internal structure of the system and its state variables. It describes only a 'black box' system. The state space method in modern cybernetics can not only describe the relationship between system input and output, but also completely describe the internal structure and the relationship between its state variables. It describes a 'white box' system. Because more system information can be described, better system control can be achieved. Cybernetics, information theory and system theory, as independent disciplines, have their own development directions and internal relations at the same time. In the study of communication and control, it is inseparable from the system; When studying system or control, information is indispensable. The general system theory considers its research object as a whole, puts forward the models, principles and laws suitable for all systems, and emphasizes that the system is for the individual, which helps to explain the organized system.