Synthesis of Optimal Control Law for Nonlinear Systems Based on Integration of Artificial Neural Network and LQR Controller

The paper sugests a method employing artificial neural network algorithms in conjunction with LQR (Linear Quadratic Regulator) control for nonlinear systems, with the objective of minimizing a quadratic performance index. When using standalone artificial neural networks, the ability to approximate the network’s function is limited by the number of neurons, making it difficult to achieve global minimum values. Combining neural networks and LQR (Linear Quadratic Regulator) is expected to enhance the training process of neural networks, making it faster and more effective. Simulation results affirm the superiority of the system combining an artificial neural network and the LQR control, compared to the use of an independent LQR control system.