Scenario-based safety validation for Connected and Automated Driving

Connected Automated Driving (CAD) technology is anticipated to be a key aspect for achieving a higher level of road safety. As CAD systems become increasingly integrated and complex, case-by-case testing of individual functions is no longer sufficient to ensure operational safety on the road. The StreetWise methodology is developed to build and maintain a real-world scenario database, suitable for testing and validation of CAD functions.

Test drives with prototype automated driving systems appear to be an enormous effort. It does not seem to be feasible to drive these millions of kilometres with the increased speed of development of automated driving functions and the high level of safety requirements that are expected from these functions. Moreover, most of the events occurring during these test drives are rather common, while other events might only occur once in several millions of kilometres. Therefore, there is an essential need for constructing and collecting relevant traffic events and situations (scenarios) for testing and validation of CAD functionalities. The collection of these scenarios should in principal represent and cover the entire range of real-world traffic situations that might be encountered by the CAD system-under-test.

In StreetWise, real-world scenarios are extracted from microscopic traffic data, i.e. data collected on the level of individual vehicles. Typically they include e.g. road layout, subjects involved, manoeuvres, relative distances, speeds, view blocking obstructions, weather and light conditions.

The StreetWise database is continuously extended by monitoring the situations that vehicles encounter on the road and distinguishing between different scenarios. Gradually, with analysing a large number of kilometres and hours of data, this collection of scenarios will be more and more representative for the situations that a car may encounter when deployed on public roads in a certain geographical area. The database stores the collected scenarios efficiently and allows for fast search of the appropriate scenarios for testing specific CAD functions.

In a Vision Paper, a detailed description of the StreetWise methodology as developed by TNO is presented. The methodology provides a solution on how to create such a scenario database from real-world data and how to use it in the assessment of automated driving functions. A roadmap is provided on how the methodology is designed to evolve with the increasing automation and communication in the road mobility systems.