Formal Scenario-driven Logical Spaces for Randomized Synthetic Data Generation

Osama Maqbool, Jürgen Roßmann

2022

Abstract

Simulations and synthetic data are a necessary supplement to real-world experiments in order to alleviate its effort, cost and risks. As demand of data for development and validation increases, simulations too must correspondingly be scaled. Variation of simulation parameters affords simulation designers control over the scope of how a simulation is scaled— they can chose a balance between target distribution of simulation variants and the degree of randomness— thereby achieving both the volume and diversity of synthetic data. This paper proposes logical scenarios as basis for simulation variation. Scenarios are formal human-readable scripts of simulations and test drives used within the automotive industry. They are defined at different abstraction levels, one of which is the logical scenario as a parameterized simulation model with description for parameters instead of concrete values. This contribution proposes methodologies to model the parameter descriptions in a modular fashion with parameter ranges, probability distributions and inter-relations. A randomization engine is introduced based on Markov chain Monte-Carlo methods to efficiently sample the modeled space. The result is a variety of simulation-independent concrete scenarios that follow the formal scenario specification.

Paper Citation

in Harvard Style

Maqbool O. and Roßmann J. (2022). Formal Scenario-driven Logical Spaces for Randomized Synthetic Data Generation. In Proceedings of the 10th International Conference on Model-Driven Engineering and Software Development - Volume 1: MODELSWARD, ISBN 978-989-758-550-0, pages 203-210. DOI: 10.5220/0010816400003119

in Bibtex Style

@conference{modelsward22,
author={Osama Maqbool and Jürgen Roßmann},
title={Formal Scenario-driven Logical Spaces for Randomized Synthetic Data Generation},
booktitle={Proceedings of the 10th International Conference on Model-Driven Engineering and Software Development - Volume 1: MODELSWARD,},
year={2022},
pages={203-210},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010816400003119},
isbn={978-989-758-550-0},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 10th International Conference on Model-Driven Engineering and Software Development - Volume 1: MODELSWARD,
TI - Formal Scenario-driven Logical Spaces for Randomized Synthetic Data Generation
SN - 978-989-758-550-0
AU - Maqbool O.
AU - Roßmann J.
PY - 2022
SP - 203
EP - 210
DO - 10.5220/0010816400003119