ABSTRACT
In this section, we explain an application of a fuzzy controller for motion control of two types of brachiation robot: a two-link brachiation robot and a seven-link brachiation robot, which move dynamically from branch to branch like a gibbon, a long-armed ape, swinging its body like a pendulum (figure G2.4.1). First, we introduce a reinforcement learning algorithm which updates the fuzzy controller based on fixed evaluation function so that the two-link brachiation robot should perform the objective tasks in computer simulations. The algorithm can deal with a range of continuous real-valued actions. The reinforcement signal is self-scaled, which prevents the variables in the fuzzy controller from overshooting when the system receives very large reinforcement values. Next, we show that a hierarchical behavior controller can control more complex robot systems. The hierarchical behavior controller enhances the capability of dealing with the larger number of input and output variables which are necessary to control the complex systems. The hierarchical control structure is applied to the control of the seven-link brachiation robot in computer simulations.
