Researchers at the Institute of Control Engineering investigate methods and concepts for the safety assurance of autonomous road vehicles. Because real-world testing alone cannot cover the vast number of relevant scenarios, simulation plays an increasingly important role. Yet every simulation model is a simplification: it represents reality only partially, with limited accuracy and only for certain conditions. Whether simulation results can be trusted therefore depends on whether the models are used within the conditions for which they are valid.
To make this explicit, the institute has developed an approach that describes each model’s validity domain as a contract: under which operating conditions (assumption) the model delivers which accuracy (guarantee). By combining the contracts of all models in a simulation setup, one can check automatically whether the setup is sufficiently valid for a given test case. A Python prototype of this approach exists. In industry, simulation models are commonly exchanged as Functional Mock-up Units (FMUs) and assembled into simulation setups using the System Structure and Parameterization (SSP) standard.
In this master’s thesis, the contract approach shall be brought to these industry standards. First, a literature review on FMUs, SSP and their extension mechanisms, as well as on contract-based design, is to be conducted. Building on an existing contract specification, an exchange format shall be developed that lets contracts travel together with FMUs and SSP setups. Subsequently, a software tool (programming in Python) shall be implemented. The tool combines the contracts along the connections of a simulation setup, checks them against the requirements of a test case, and monitors them while the simulation runs. A proof of concept shall demonstrate the tool on an existing case study and one further example, followed by a critical evaluation. The thesis is part of a work package with a second master’s thesis; both can be worked on independently, and collaboration is welcome.
Required are...
Helpful, but not strictly required are...