In daily life, the so-called adaptive means a feature that biore can change its habits to accommodate a new environment. Therefore, intuitively, the adaptive controller should be such a controller that corrects its own characteristics to adapt to changes in dynamic characteristics of objects and disturbances. The research object of adaptive control is a system having a certain degree of uncertainty. The so-called "uncertainty" refers to the mathematical model described in describing the controlled object and its environment, which contains some unknown factors and random factors. Any actual system has different degrees of uncertainty, which sometimes manifests inside the system, sometimes manifested outside the system. From the inside of the system, the structure and parameters of the mathematical model of the controlled object are described, and the designer does not necessarily know accurately in advance. As an external environment affects the system, it can be denoted by many disturbances. These disturbances are usually unpredictable. In addition, there are some uncertain factors generated when measurement enter the system. In the face of these objective existence, how to design proper control, so that a designated performance indicator reaches and maintains optimal or approximate optimal, this is the problem of adaptive control to study solutions. The same is true for adaptive control and conventional feedback control and optimal control, and a control method based on mathematical models, which is only relatively small, and the prior art of the model and disturbance is relatively small, requiring the operation of the system. In the process, you will continue to extract information about the model, so that the model is gradually improved. Specifically, the model parameters can be constantly identified according to the input of the object, which is called the system's online recognition. With the continuous production process, through online recognition, the model will become more accurate, getting closer to the actual. Since the model is constantly improving, it is clear that the control role integrated based on this model will continue to improve. In this sense, the control system has certain adaptability. For example, when the system is in the design phase, due to the lack of the initial information of the object characteristics, the system may not be ideal when it starts to put the runtime, but as long as the operation is running, after online recognition and control, the control system is gradually adapted. The final will eventually adjust itself to a satisfactory working state. For example, some control objects, its characteristics may have changed during operation, but the system can be gradually adapted by online recognition and changing the controller parameters. Conventional feedback control systems have certain suppression capabilities for changes in the internal characteristics of the system and external disturbances, but since the controller parameters are fixed, the system's internal characteristics change or the amplitude of the external disturbance is large. Performance often decreases, even unstable. So the variation of object characteristics or disturbance characteristics is large, and it is necessary to maintain a class of adaptive controls that often maintain high performance indicators. It should also be pointed out that adaptive control is much more complicated than conventional feedback control, and there is only a high cost, so it is only considered when it is not expected with conventional feedback.
