A LQR-based Neural-network Controller for Fast Stabilizing Rotary Inverted Pendulum

Inverted Pendulum (IP) is an unstable and highly nonlinear system. It is used as a common model for engineering applications in linear and nonlinear control. So, IP has been frequently listed as an ideal showcase for different controller models. In this paper, we present an advanced control method for stabilizing the IP system in upright posture. The key technology is a combination of Linear Quadratic Regulator (LQR) and an online Radial Basis Function (RBF) Neural-Network compensator. The LQR term plays a dominant role in yielding the nominal control signal based on a linearized model, while the RBF term is employed to tackle the systematic deviation as the system is far from the equilibrium state. A damping segment-wise learning law is proposed to activate the network operation. Theoretical analyses and comparative validation results provided success evidences of the designed controller.