Anomaly detection using the Nash equilibrium in a multi-agent system

In order to support operators of surveillance systems and to improve their situation awareness, the automatic detection of unusual behavior and anomalies can provide an essential contribution. As the amount of available data increases due to cheaper and more proficient sensor systems like video cameras and radar systems, a human operator might not be able to cope with it. This might result in a misinterpreted situation and ultimately in making wrong decisions. Therefore, this report deals with an anomaly detection algorithm based on a multi-agent system. Each object in the monitored area is modeled as an agent with its own utility function. The agents are supposed to behave in a rational way. As the Nash equilibrium is a game theoretic concept to utilize rational behavior, the approach in this report is to use this solution concept to model the agents' behavior.