A study to model human behavior in Discrete Event Simulation (DES) using Simkit

Abstract

In modern simulation systems, there are two main domains. The first is the Event Driven domain that uses events as the driver in the simulation. Event driven simulations are efficient as they avoid unnecessary time steps and, when carefully designed, can realistically represent real world events. The Time Driven paradigm, the other simulation domain, is favored by agent-based simulations due to the time stepping characteristic suitable for the agent decision cycle. Agent-based simulations such as the multi-agent systems are capable of modeling complex human intelligence and behavior. In this thesis, Discrete Event Multi Agent Simulation (DEMAS) is introduced as a new design concept which provides the mean to have the best of both simulation domains. DEMAS design concept uses event graph methodology and LEGO framework to achieve design simplicity and modularity, allowing multi-agent systems to be added into a discrete event world. To validate the new design concept, three simulations of different complexity levels were developed using Simkit Java package. The validations eventually proved the worthiness of the DEMAS design concept for providing the means to build simulations with benefiting characteristics of both multi-agent systems and discrete event simulations.