On-the-fly fluid model checking via discrete time population models

We show that, under suitable convergence and scaling conditions, fluid model checking bounded CSL formulas on selected individuals in a continuous large population model can be approximated by checking equivalent bounded PCTL formulas on corresponding objects in a discrete time, time synchronous Markov population model, using an on-the-fly mean field approach. The proposed technique is applied to a benchmark epidemic model and a client-server case study showing promising results also for the challenging case of nested formulas with time dependent truth values. The on-the-fly results are compared to those obtained via global fluid model checking and statistical model-checking.