Scenario-Based Testing of Autonomous Vehicles: The Foundation of Safety and Reliability

Marek Mičák

---

The development of autonomous vehicles and Advanced Driver Assistance Systems (ADAS) creates a fundamental need to ensure their safe operation in real traffic conditions. Scenario-based testing has become a key tool that enables the systematic simulation of various traffic situations and the verification of system responses before deployment in real-world traffic.[1] This approach is also supported by regulatory authorities, as modern regulations increasingly require the validation of autonomous system safety through predefined scenarios.[2] UN[3] and EU[4] regulations rely on scenario-based testing to assess the safety of automated driving systems and include scenario specifications and acceptance conditions. Although these predefined scenarios cover important safety aspects, the development of safe autonomous vehicles requires additional testing, placing several demands on test scenarios and their generation process.[5]

Scenario-based testing makes it possible to reproduce real-world traffic situations, including critical and rare events that would be dangerous to test in real conditions. Such scenarios include, for example, collision situations, sudden changes in the behaviour of road users, or complex interactions at intersections. The advantage is that these scenarios can be repeatedly tested in a controlled environment, increasing the accuracy of analysis and enabling the identification of errors already in the early stages of development.

Modern testing of autonomous vehicles is no longer based solely on physical testing but on a comprehensive ecosystem that combines simulations, proving ground testing, and real-world road testing. This hybrid approach allows efficient coverage of a wide range of scenarios while optimizing development time and costs. Virtual simulations are particularly important for testing extreme or difficult-to-reproduce situations, which can then serve as a basis for physical validation.[6]

From a scientific perspective, several approaches exist for generating test scenarios. The main ones include data-driven scenarios, scenarios generated by artificial intelligence, and knowledge-based scenarios derived from traffic rules or formal specifications. Each of these approaches has its advantages and limitations—data-driven scenarios are realistic but limited by available data, while AI-generated scenarios offer greater variability but less control.

To overcome these limitations, integrated solutions have been developed that combine various scenario generation techniques into a single system. Such an approach makes it possible to create realistic, controllable, and critical scenarios, significantly increasing testing efficiency. An important feature of modern tools is their modular architecture, which allows flexible adaptation of testing processes to the specific needs of autonomous system development.

Simulations play a crucial role, particularly in terms of efficiency. They enable the generation of large numbers of scenarios in a short time and allow testing of vehicle behaviour under various traffic conditions. This approach significantly accelerates development and enables extensive testing without the need for costly physical experiments.[7]

At the same time, it is essential to combine virtual testing with physical validation. While simulations can identify potential issues, real-world testing is necessary to confirm results under actual conditions. This hybrid approach ensures that testing is not only efficient but also reliable and relevant for real-world deployment.[8]

Scenario-based testing therefore represents a fundamental pillar in the development of autonomous vehicles. It enables systematic evaluation of system behaviour across different situations, identification of weaknesses, and compliance with regulatory requirements. As the complexity of autonomous technologies continues to grow, its importance will further increase and will be a key factor in building public trust in autonomous mobility.[9]

 

 

SOURCES:

[1] TÜV SÜD – Scenario-based testing for automotives. In: TÜV SÜD. [online]. [cit. 24.03.2026].  Dostupné na: https://www.tuvsud.com/en/industries/automotive/autonomous-vehicle-testing-and-homologation-services/scenario-based-testing-for-automotives

[2] Finkeldei, F., Thees, C., Weghorn, JN. et al. Scenario Factory 2.0: Scenario-Based Testing of Automated Vehicles with CommonRoad. Automot. Innov. 8, 207–220 (2025).  [online]. [cit. 24.03.2026].  Dostupné na: https://doi.org/10.1007/s42154-025-00360-0

[3] World Forum for Harmonization of Vehicle Regulations: New Assessment/Test Method for Automated Driving (NATM) Guide- lines for Validating Automated Driving System (ADS). United Nations, Geneva (2023)

[4]  European Comission: Commission implementing regulation (EU) 2022/1426. Official Journal of the European Union 65, 1–64 (2022)

 

---