Improving simulation specification with MBSE for better simulation validation and reuse

A simulation can be a complex architecture of simulation models, simulation tools, and computing hardware. However, its development often relies on informal procedures and can begin without a clear, complete, and formal definition of the simulation needs. Simulation traceability is then compromised, which prevents from easily validating whether a simulation meets the needs, or understanding the purpose of a simulation model that can be reused. This paper proposes an approach to improve the definition of simulation needs using Model-Based Systems Engineering. Based on the semi-automatic processing of a system architecture, it presents a new method to formulate a so-called "simulation request" which covers (1) the part of the system to be simulated; (2) the objective of the simulation; (3) the simulation quality, cost, and delivery; (4) the test scenarios; (5) the data for simulation calibration and validation; and (6) the verification and validation of the simulation. All the tooling required for the formulation of the simulation request were prototyped in a SysML editor, with machine learning capabilities for the choice of test scenarios. The method and tooling were tested for the case of an autonomous car passing under traffic lights.