This paper discusses a new architecture for accelerator tuning that combines
heuristic and knowledge based methods with traditional approaches to control.
Control of particle accelerators requires a hybrid architecture, which
includes methodologies for planning, intelligent search, and pattern
recognition. Control is distributed and hierarchical to utilize parallel
problem-solving in the face of time-sensitive control requirements and to
decompose complex control problems into more manageable subtasks. For
perspective, we discuss past attempts at accelerator control and why these
attempts left many issues unresolved. We describe the details of our
control architecture along with its motivation. We then report the results
of deploying and testing it at two accelerator facilities. This paper ends
with a discussion of the commercial importance of this work.
To be published in the Proceedings of the Innovative Applications of
Artificial Intelligence 1997 by Klein,
W., Stern, C., Kroupa, M., Westervelt, R., Luger, G., and Olsson, E.
[Intelligent Control][Home]