• E. Meier, G. LeBlanc
  ASCo, Clayton, Victoria
• S. Biedron
  ELETTRA, Basovizza
• M.J. Morgan
  Monash University, Faculty of Science, Victoria

We report the development of an artificial intelligent system for the optimisation of electron beam parameters at the Australian Synchrotron Linac. The system is based on state of the art developments in Artificial Intelligence techniques for video games and is adapted here to beam parameters optimisation problems. It consists of a genetically evolved neural network that mimics an operator's decisions to perform an optimisation task when no prior knowledge other than constraints on the actuators is available. The system's decisions are based on the actuators positions, the past performance of close points in the search space and the probability of reaching a better performance in the local region of the search space.