Self-Tuning State-Feedback Control of Rotary Pendulum via Online Adaptive Reconfiguration of Control Penalty-Factor

Abstract

This paper presents an online self-tuning state-feedback controller for under-actuated systems in order to enhance the system’s position-regulation accuracy, disturbance-rejection capability and control input efficiency. The Linear-Quadratic-Regulator (LQR) is employed as the baseline state-feedback controller whose gains are dynamically reconfigured by adaptively modulating the control penalty-factor associated with the quadratic performance-criterion associated with the controller. The main contribution of this article is the methodical formulation of an original reconfiguration block that dynamically adjusts the control penalty-factor via a self-regulating secant-hyperbolic function to flexibly manipulate the control procedure while maintaining its asymptotic-stability. The proposed reconfiguration strategy modifies the control penalty-factor online with respect to the state-error feedback as well as the control-input dynamics. The benefits afforded by the proposed self-tuning controller are verified by conducting hardware-in-the-loop experiments on the QNET 2.0 Rotary Pendulum board. The experimental results show that the proposed strategy renders rapid transits with strong damping against parametric uncertainties in pendulum’s body-angle responses, while limiting the peak servo requirements of the actuator.