Discretization of Parameter Space for Virtual Scenario Based Testing

Masterthesis, Bachelorthesis

The Institute of Automotive Engineering (FZD) is involved in the project Virtual Toolchain for Automated and Connected Driving (VIVALDI). Scenario based testing has emerged as a promising approach for the safety assurance of automated driving. One objective is the development of a methodology to find an adequate discretization of the parameter space for a given scenario.

Finding a suitable discretization step for each parameter is crucial to ensure efficient coverage of the parameter space while maintaining a sufficiently low discretization error. Current approaches for discretization of scenario parameters rely on heuristics and lack quantification of the error. Therefore, an attempt to adapt methods from domains such as computational fluid dynamics (CFD), Metamodeling and Design of Experiments (DoE) is required.

Tasks include:

  • Research on discretization and error estimates in other domains
  • Evaluate the applicability to the problem of scenario parameter discretization
  • Develop method for setting the discretization step
  • Develop method for estimating the discretization error
  • Extend the existing toolchain for simulation with perception evaluation
  • Implementation of methods for discretization and error estimation

Requirements:

  • Experience with Python and CarMaker may be beneficial, but not required
  • Knowledge of discretization from other domains like structure or fluid dynamics recommended
  • Analytical mindset, problem solving aptitude and structured working style