Towards the Verification of Hybrid Co-simulation Algorithms

摘要

Engineering modern systems is becoming increasingly difficult due to the heterogeneity between different subsystems. Modelling and simulation techniques have traditionally been used to tackle complexity, but with increasing heterogeneity of the subsystems, it becomes impossible to find appropriate modelling languages and tools to specify and analyse the system as a whole. Co-simulation is a technique to combine multiple models and their simulators in order to analyse the behaviour of the whole system over time. Past research, however, has shown that the naive combination of simulators can easily lead to incorrect simulation results, especially when co-simulating hybrid systems. This paper shows: (ⅰ) how co-simulation of a family of hybrid systems can fail to reproduce the order of events that should have occurred (event ordering); (ⅱ) how to prove that a co-simulation algorithm is correct (w.r.t. event ordering), and if it is incorrect, how to obtain a counterexample; and (ⅲ) how to correct an incorrect co-simulation algorithm. We apply the above method to two well known co-simulation algorithms used with the FMI Standard, and we show that one of them is incorrect for the family of hybrid systems under study, under the restrictions of the standard. The conclusion is that either the standard needs to be revised, or one of the algorithms should be avoided.