Modeling through model transformation with MARS 2.0

The development of simulation models confronts scientists with the necessity of transforming concepts from theoretical models to executable code. Albeit modern simulation platforms provide APIs to abstract away technology, this task remains complex. Therefore a model-to-code transformation is essential, allowing domain experts to focus on their model instead of implementation details. This paper presents a multi-level transformation concept to facilitate building multi-agent simulations for domain experts. With domainspecific tool support, model ideas can be developed without managing technical requirements. This insures that the modeler is exclusively concerned with the conceptual model by utilizing a MARS (Multi-Agent Research and Simulation) Meta-Model (MMM) and Agent Meta-Model (AMM). We outline the MMM as underlying structure, discuss the foundations of model-driven development and the in-place transformation of the MMM as executed by the MARS modeling workflow. In addition, we present a model-to-code generator that creates the final simulation model.