QLQG/LTR control of parallel inverted pendulums

Abstract

The parallel inverted pendulum system is one of the representative mechanical systems that have inherent instability characteristics, geometrical nonlinearities and dry frictions in actuators. Hence, it is sometimes desirable to apply a nonlinear multivariable control methodology. In the paper, a nonlinear quasi-linear quadratic Gaussian with loop transfer recovery (QLQG/LTR) control is proposed and experimented with on a parallel inverted pendulum system with dry frictions, in which system parameters such as amp gains are identified using genetic algorithms, and the dry friction is approximated to a quasi-linear model by random input describing function techniques.