Exploiting Structural Dependency Relations for Efficient Agent Based Model Simulation

In the last few years Agent Based Models (ABMs) have attracted growing interest in the field of computational simulation thanks to their applicability in very heterogeneous landscapes, usability for fine-grained descriptions and comprehensibility for application domain experts. However, the lack of a well-defined semantics for specifying how agents behave and how they get coupled and scheduled may lead to inconsistent results. To fill this gap we proposed a well defined ABMs semantics that, using Extended Stochastic Symmetric Nets for model description, allows the modeller to automatically derive the corresponding ABM simulator that is directly executable in the NetLogo ABM framework. In the present paper we propose an improvement that exploits locality of state change effects to avoid recomputing the rates of the enabled events at each state change. This is achieved by exploiting structural properties of the ESSN model to generate optimized NetLogo code (semi)automatically. The results obtained for an example case-study demonstrate a relevant improvement in terms of execution time when structural optimizations are employed to reduce rates calculations.