MOOVE Project: Recognition of Road Scenes by the Data Collected at the Output of the Sensors of the Autonomous Vehicle

Autoren

Dr. Annie Bracquemond, Gildas Thiolon, VEDECOM, Versailles

Zusammenfassung

Knowledge and modelling of the driving environment is fundamental to learn and understand complex road situations, as well as identifying the critical safety scenarios that will be encountered by delegated driving vehicles.

The assumption of every global car manufacturer is that the only way to identify, and therefore deal with the complexity and diversity of traffic situations on our roads is to collect vast amounts of data (about 1 million kilometers of actual data). This collection is made from a fleet of vehicles equipped with sensors and driven by professional drivers strictly respecting traffic regulations and the rules of road safety, as an automated vehicle would do.

By building a real-world road traffic scenario database (DB) that serves as a technical reference for the design of automated delegated driving functions, the MOOVE project led by the Vedecom Institute identifies the most relevant driving scenarios. These will be used in simulation to validate new automated driving functions, especially those related to safety (safety).

This means that algorithms and the overall operation of autonomous driving must be validated using a "Big Data" approach. From these data, driving scenarios and their most relevant environmental factors are extracted and modelled. The extraction of these scenarios requires the development of a database of rules for interpreting these scenarios in order to produce "expert system" type algorithms.

Mitglieder-Login