MEDART-MAS: MEta-model of Data Assimilation on Real-Time Multi-Agent Simulation

Abstract

In modeling and simulation process, data plays an important role. Data is required to validate the model and to experiment scenarios. It is also necessary for fitting and calibrating model parameters. In the case of online simulation, data assimilation approaches make possible to inject data into simulations and to recalibrate simulations based on real-time data. This paper addresses the challenge of assimilating data into an agent-based simulation by promoting a novel architecture dedicated to data assimilation. Few improvements have been made to adapt Multi-Agent Simulations to real-time data assimilation. The architecture is designed to be generic enough to allow wild diversity of case studies. We propose a meta-model of data assimilation and implement a toolkit based on the GAMA simulator. Finally, we use temperature data to test the implementation of a simple use case.