Control of the Acrobot with Motors of Atypical Size Using Artificial Intelligence Techniques

Abstract

An acrobot is a planar robot with a passive actuator in its first joint. The main purpose of this system is to make it rise from the rest position to the inverted pendulum position. This control problem can be divided in the swing-up issue, when the robot has to rise itself by swinging up as a human acrobat does, and the balancing issue, when the robot has to maintain itself in the inverted pendulum position. We have developed three controllers for the swing-up problem applied to two types of motors: small and large. For small motors, we used the State-Action-Reward-State-Action (SARSA) controller and the proportional–derivative (PD) controller with a trajectory generator. For large motors, we propose a new controller to control the acrobot—a pulse-width modulation (PWM) controller. All controllers except SARSA are tuned using a genetic algorithm.

Keywords

• acrobot
• artificial intelligence
• SARSA
• PWM
• genetic algorithm