Leveraging Regular Expressions for Flexible Scenario Detection in Recorded Driving Data

Abstract

For the development of Advanced Driver Assistant Systems (ADAS) and Automated Driving Systems (ADS) large amounts of driving data are collected. Storing, processing and analyzing this data requires significant efforts. For these efforts to pay off, data must be made accessible and reusable for different development phases from specification to testing. However, raw measurement data can largely vary in format, configuration and comprehensiveness. Adding meaningful, semantic labels to the data, also called data enrichment, is a promising way to improve searchability and to compute aggregating statistics over large databases. In this work we identify challenges and define requirements for data enrichment of real world measurement data for ADAS and ADS. We suggest a framework as a general and modular processing pipeline to tackle common challenges of real world data. Within this framework we support Boolean expressions to reduce multivariate time series to states of interest. Finally regular expressions are used as a deterministic method to identify semantically meaningful sequences. We demonstrate the flexibility of this approach by detecting lane change and cut-in maneuvers. The proposed methods prove to be well suited for robust maneuver detection with varying durations.

Topics & Concepts