Learning-based adaptive control method

The invention discloses a learning-based adaptive control method, and relates to the technical field of control engineering. The control method comprises the following steps: step 1, a self-adaptive control method based on extremum search: firstly, designing a performance cost function, measuring the cost function and the gradient thereof by using an extremum search algorithm, generating a series of expected states which can enable a controlled system to reach the optimal value of the performance cost function, then designing a model-based controller to enable the state of the system to track the expected states; and finally, proving the convergence of the method, namely it can be guaranteed that the system continuously operates near the ideal performance function extreme value. According to the method, for a control object with uncertainty, the uncertainty is taken as an entry point, an appropriate model-based controller design method and a model-free learning algorithm are selected, and finally the advantag