An Intelligent Control Architecture for nin
William B. Klein, Carl R. Stern, George F. Luger, Eric T. Blsson · 1997
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. The Accelerator Control Broble Tuning particle accelerators is time consuming and expensive, with a number of inherently non-linear interactions between components of the system. Conventional control methods have not been successful in this domain, and the result is constant and expensive monitoring of the systems by human operators. In recent years with isolated successes, advanced technologies such as expert systems, neural networks, and genetic algorithms have been applied to the individual pieces of this problem.