Gregory Pavlov

Computer Science Advisor: Prof. Robert Signorile Intelligent Entities in a Distributed Simulation Environment Abstract In addressing the growing model sizes used in simulation environments, I examine adding machine learning techniques to the entities in a model in an effort to produce such side effects as emergent behavior. Distributing the environment in order to increase the efficiency of the simulation also plays an important role in this thesis. The added intelligence of entities may have some affect upon the speed at which a model may be executed, as additional computation will be required. Taking a model-based approach, I attempt to solve some of the problems of interaction between components in a distributed simulation. Introduction Simulation is the product of science attempting to model the real world as accurately as possible in a virtual world. The goal being to make predictions regarding the result of a given decision or choice, and then see the effects upon the created realm, as the first computers were used to compute artillery shell trajectories (Fujimoto 7). As the required workload given to a specific program operating on a single machine began to grow such that a simulation could no longer run in real-time, breaking up the model and sharing the work among multiple machines became much more common. The design of each system used the modules to order the entities to different locations in the model, rather than the entities making a certain choice when required. Discrete-event simulation uses modules to perform some task or function on an entity in the model upon arrival at the given module. The model for such a simulation appears similar to a workflow diagram, where some modules create entities, others delay, others decide the path a