Recursive-gradient-based Model-Free Adaptive Control with RBF Neural Network

Model-free adaptive control (MFAC) is the so-called data-driven control method which mainly uses the I/O data to estimate the pseudo partial derivative (PPD) of the system online, and then use the corresponding estimated PPD to obtain the control law. In order to reduce the fluctuation of the loss function and accelerate the speed of system convergence, the improved recursive-gradient-based MFAC algorithm is proposed in this paper. Because the PPD not only denotes the parameter but also represents the structure of the system, the inaccurate estimation of PPD may lead to instability when the controlled system is too complex. By taking the ability of good approximation for nonlinear function and the rapid convergence speed into consideration, the RBF neural network (RBFNN) is applied to estimate the PPD online in our proposed recursive-gradient-based MFAC algorithm. The simulation results show that the control accuracy and the convergence speed have been improved which verify the effectiveness and feasibility of the proposed algorithm.