Adaptive swing-up and balancing control of acrobot systems

Abstract

The field of underactuated robotics has become the core of agile mobile robotics research. Significant past effort has been put into understanding the swing-up control of the acrobot system. This thesis implements an online, adaptive swing-up and balancing controller with no previous knowledge of the system's mass or geometric parameters. A least squares method is used to identify the 5 parameters necessary to completely characterize acrobot dynamics. Swing up is accomplished using partial feedback linearization and a pump up strategy to add energy to the system. The controller then catches the swung up system in the basin of attraction of an LQR controller computed using the estimated parameter values generated from online system identification. These results are then simulated using a MATLAB simulation environment.