Autonomous Agent-Based Simulation of a Model Simulating the Human Air-Threat Assessment Process [open pdf - 2MB]
The Air Defense Laboratory (ADL) Simulation is a software program that models the way an air-defense officer thinks in the threat assessment process. The model uses multi-agent system (MAS) technology and is implemented in Java programming language. This research is a portion of Red Intent Project whose goal is to ultimately implement a model to predict the intent of any given track in the environment. For any air track in the simulation, two sets of agents are created, one for controlling track actions and one for predicting its identity and intent based on information received from track, the geopolitical situation and intelligence. The simulation is also capable of identifying coordinated actions between air tracks. Three kinds of aircraft behavior in the simulation were used: civilian, friendly and enemy. Predictor agents are constructed in a layered structure and use "conceptual blending" in their decision-making processes using mental spaces and integration networks. Mental spaces are connected to each other via connectors and connectors trigger tickets. Connectors and Tickets were implemented using the Connector-based Multi Agent System (CMAS) library. This thesis shows that the advances of Cognitive Science and Linguistics can be used to make our software more cognitive. This simulation is one of the first applications to use cognitive blending theory for a military application. It was demonstrated that agents can create an "integration network" composed of "mental spaces" and retrieve any mental space data inside the network immediately without traversing the entire network by using the CMAS library.
Naval Postgraduate School, Dudley Knox Library: http://www.nps.edu/Library/index.aspx