Adaptive Neural Control for a Two Wheeled Inverted Pendulum System

This paper presents an adaptive neural network (ANN) based control system to stabilize a two wheeled inverted pendulum (TWIP) system which is an underactuated and highly unstable system with a complex nonlinear dynamics. The proposed ANN scheme is composed by two adaptive layers that are trained online to identify the dynamics of the TWIP system by means of the augmented backpropagation algorithm. This uses novel adaptive laws based on adaptive control techniques in order to robustify control performance. The real-time experiment shows the behavior of the proposed controller on the system "RAB-ITLAG-1". The performance criterion for the obtained control responses is given by the root mean square (RMS) value.