Two Decades of Adaptive Control Pitfalls
<p>
 Brian D.O. Anderson
 <p>Chief Scientist, National ICT Australia Limited
 and Distinguished Professor, Australian National University
<p>
 Adaptive control is a very appealing technology, at least in
 principle.  Yet its use has been conditioned by an attitude of
 distrustfulness on the part of some practitioners.  In this talk, we
 review (without presupposing knowledge of adaptive control on the
 side of listeners) some concepts the isolation of which was necessary
 to engender confidence in the technology.  These include the
 unpredictable failure of the MIT rule; the bursting phenomenon, and
 how to prevent it; the notion that identification of a plant is only
 valid conceptually for a restricted range of controllers (with the
 implication that in adaptive control, certain controller changes may
 be hazardous); and the concept of multiple model adaptive control,
 which from some points of view raises as many the problems as it solves.