The inverted wedge is a planar robot with two degrees of freedom and a sing
le control input (i.e., it is "underactuated"). The goal is to develop a di
gital controller that can balance the wedge in the inverted position by shi
fting a weight on the top of the wedge. Because it is underactuated and has
complicated nonlinear dynamics. the inverted wedge is a good testbed for t
he development of nonconventional advanced control techniques and comparati
ve analysis between control methods. We begin with the development of a non
linear state feedback controller and direct and adaptive fuzzy controllers.
that we will later use as a baseline comparison to show what type of perfo
rmance is possible for this testbed. Control routines based on the GA have
been found to apply to several practical applications in simulation and off
-line optimization. Here, we will show that a GA can be used on-line in rea
l-time to produce a particularly effective adaptive control method and this
is the main contribution of this work. Computational and real-time impleme
ntation issues will be discussed and the genetic adaptive strategy will be
compared with the state feedback and fuzzy control methods. (C) 2001 Elsevi
er Science Ltd. All rights reserved.