Neural network-based model reference control for inverted pendulum

In this paper, we stabilize an inverted pendulum using neural network controller. The inverted pendulum is a nonlinear single-input double-output system. One of the outputs is the pendulum angle, another is the cart position which carries the inverted pendulum. We propose a new control method for such a controlled object. This neural network controller is mainly composed of two parallel I-PD compensators. The performance of the proposed controller is determined by its control gains. However, it is difficult to tune these gains manually to achieve the desired dynamic response. To adjust these gains, we use multilayer neural networks including the sigmoidal functions. By using the sigmoidal functions, we can construct the nonlinear controller. We also show the effectiveness of the proposed neural network controller.