A Novel Data-Driven Hybrid Adaptive Control

As complementary methods to classical model-based control, data-driven control methods can avoid modeling the dynamics of complex systems but achieve good control performance, hence, data-driven control methods are gradually becoming an important and hot topic in control science and engineering. However, some issues in data-driven control methods still need to be further addressed such as physical interpretability, excessive hyperparameters, control parameter valid initialization, etc. In this paper, a novel algorithm namely data-driven hybrid adaptive control (DHAC) is proposed, which fully integrates parameter tuning theory and artificial neural network theory. Specifically, DHAC algorithm as a model-free method has the advantages of classical control methods and adaptive neural networks methods, while eliminating the drawbacks of both two types of methods. Hence, DHAC algorithm has the potential to provide new method to solve the issues of control parameters valid initializations, physical interpretability, excessive hyperparameters in the area of model free control.