VOMAS for Validation of Agent-based Models - Requirements and Application

Abstract

Agent-based models are difficult to validate. This is mostly due to their complex structure and emergend behavior. Therefore, new validation methods targeted at agent-based models are developed. One of those methods uses a virtual overlay multi-agent system (VOMAS), which is able to monitor the simulation and its agents, check for violations of model assumptions and if necessary can log information for later use. In this paper we present a detailed concept of VOMAS components with a platform-independent design, which offers easy and flexible usability on part of the user. To test the VOMAS design, it is applied to an existing agent-based model. The developed VOMAS proves to be a very flexible tool for validating agent-based models, allowing the user to choose which simulations agents should be observed and what attributes need to be monitored and logged. Applied to the test case model, the VOMAS is able to gather new information giving a detailed insight into the agent behavior that can be used to further validate the model at hand.