Investigating the implementation of knowledge representation in the COMBATXXI system

Abstract

Combat models and simulations aim to find a balance between complexity and simplicity: Both oversimplification and too much detail can lead to erroneous findings. In simulations that require representation of human behavior, modelers rely on prior scripting to find the balance. However, this technique cannot depict dynamic behavior during the simulation run. This inadequate representation of entity behavior can cause misleading or incomplete results. This thesis investigates the implementation of knowledge representation in combat models in order to enhance entity behavior. The new method does not try to include more details in the model than the scripting method, but it tries to enhance the entity decision making to create more realistic outcomes. A knowledge base along with reasoning capabilities was linked to a combat model to mimic the memory and the brain of an entity. To demonstrate the feasibility of this approach, an ontology development tool called Protégé was linked to a combat model called COMBATXXI. Besides achieving dynamic behavior, the new method has other advantages over previous approaches such as better separation of specification and implementation, loosely-coupled components to allow code reuse, use of well-established reasoners for free, and exploitation of partially-sensed information.