Back to ASAM International Conference 2019

Scenario-Based Validation Framework for highly automated/autonomous vehicles 

Hannes Schneider
Senior Software Development Engineer
AVL List GmbH

The automotive industry faces the enormous challenge of proving that highly automated and/or autonomous vehicles are significantly safer than human-driven vehicles.

The greatest difficulty is the identification of all possible traffic situations in which the vehicle must be evaluated and checked to provide this proof. There are numerous studies which, based on statistical considerations, show that a very high mileage on the road would be necessary for this proof.

In practice, the complete Validation of vehicles on the road is economically not feasible due to limited resources and time.

A promising solution to perform all necessary test scenarios is scenario-based virtual Validation. This involves extracting a large number of relevant scenarios from existing Measurement data from real test drives. Subsequently, concrete test cases are generated by varying parameters of the scenarios, which are then executed in different test environments.

This paper presents a Validation framework for automated scenario extraction from different data sources as well as the automated execution of generated test cases in different test environments - MIL / SIL / HIL / VIL (Vehicle-in-the-Loop).

The main advantage of this Validation framework is that the identified scenarios and test cases can be equally defined, automatically executed and evaluated within each test environment, supported by a high degree of standardization.

Due to the comparability of the results from different test environments, the correlation between real and virtual Measurement results is possible and form the basis for a realistic, complete Validation of highly automated/autonomous vehicles.