System design and evaluation using discrete event simulation with AI

During system design and evaluation, development of optimal decision making rules for an intelligent system can be very difficult. The Operational Evaluation Modeling (OpEM) Fuzzy, Adaptive Expert System Controller, integrated into the OpEM Simulation Tool Kit, has greatly facilitated this task. The rule specification language for Controller consists of all primitives needed to implement local and global decision making. These primitives include fuzzy facts to transform analog variables into discrete concepts and variable instantiation of facts to consider lists of facts in context-sensitive, global decisions. A single-track railroad system is evaluated as an example context-sensitive system, and other methods of decision making are compared with OpEM. Future research is discussed that adds a machine learning module to Controller, called Supervisor, to automatically search for better decision making rules.