The HolmeS3 project was set up to tackle the growing challenges of safeguarding automated driving functions. With the increasing complexity of modern vehicles and the growing demand for autonomous driving technologies, the need for comprehensive and reliable test methods is also growing. Traditional testing approaches are reaching their limits as they are not able to cover the extremely high number of possible scenarios that can occur in real road traffic.

HolmeS3 is an R&D project of the project partners Ostbayerische Technische Hochschule, e:fs TechHub GmbH and imbus AG. It was carried out from October 2020 to December 2023 and was funded by the Bavarian State Ministry of Economic Affairs, Regional Development and Energy as part of the Artificial Intelligence - Autonomous Mobility initiative.

The aim of the HolmeS3 project was to develop an improved methodology that enables the targeted and efficient safeguarding of highly automated driving functions and autonomous systems through the use of causal models.

Causal models help to understand the complex interactions between various factors in the vehicle and its environment and to systematically identify foreseeable risks. By applying these models, the test process for automated driving functions should not only be optimized, but also made more transparent and comprehensible.

The approach developed in HolmeS3 combines causal models with scenario-based testing and thus enables critical scenarios to be covered more specifically with relevant test cases. The HolmeS3 methodology thus supports causality-driven development in compliance with the requirements of ISO 21448 (SOTIF).

Click here for a detailed project report

Causal inference: Precise analysis of cause-and-effect relationships

Causal inference is a causal modeling tool that makes it possible to precisely analyze complex cause-and-effect relationships. By applying causal inference, development teams can find out how changes in one parameter variable, such as vehicle speed or a weather condition, have a direct impact on other variables, such as braking performance. Causal inference goes beyond simple correlations and allows targeted interventions to be simulated in order to optimize the robustness and safety of driving functions. In the context of ISO 21448 (SOTIF), causal inference plays a key role in identifying critical scenarios and ensuring system safety by helping to generate the most relevant test cases and avoid unnecessary test case explosions.

Generation of corner cases

A key advantage and benefit of causal modeling is the ability to identify so-called corner cases. Corner cases are special combinations of values of parameters of a causal model that represent potentially dangerous situations or states of the modeled system. Causal inference can be used to systematically generate such critical parameter value combinations from a causal model.

 

Reduction of testing effort through the use of corner cases

Using the generated corner cases as test case data makes it possible to significantly reduce the testing effort:

On the one hand, a single corner case can replace several other value combinations and thus reduce the number of test data combinations to be considered. On the other hand, the corner cases can be prioritized based on the causal model and other additional information from the causal inference, which allows the test to be further focused on the most risky test data combinations.

The solution for scenario-based testing developed in the HolmeS3 project (based on the imbus test management tool TestBench) enables the management and use of causal models and corner cases in the test process. The solution makes it possible and easier to select the appropriate corner cases for a specific driving scenario in a targeted manner and then execute the scenario with these values in the simulator system. In this way, the behavior of the driving function to be tested can be checked and validated under the extreme conditions represented by the corner cases. The result is a complete, comprehensible test documentation that contains all the information required for the safety argumentation.

Application examples and practical implementation of scenario-based testing

Scenario-based testing is widely used in the development and validation of highly automated driving functions. A typical application example is the validation of an emergency brake assist (EBA). In a simple scenario, a vehicle could follow another vehicle on a straight road, with the emergency brake assistant continuously monitoring the speed and distance. The scenario is then run with varying parameters such as different weather conditions, speeds and traffic densities to ensure that the assistant works reliably, even under extreme conditions.

Another example is the testing of automated driving functions in more complex environments, such as a busy intersection in the city center. Here, multiple factors such as pedestrians, traffic lights and sudden obstacles are integrated into the scenario. These realistic test scenarios make it possible to evaluate and optimize the vehicle's behaviour under changing and potentially dangerous conditions.

In practical implementation, a combination of simulations and real driving tests is often used. By using causal models in test management, scenarios can be precisely controlled and the most relevant parameter combinations can be tested to ensure vehicle performance in critical situations. This enables developers to carry out targeted tests that both increase safety and optimize the development effort.

Conclusion: Future-proof safety through innovative test methods

Scenario-based testing in combination with causal modeling and causal inference offers a pioneering method for validating and safeguarding highly automated driving functions. These innovative approaches make it possible to test even complex systems safely and efficiently by precisely simulating real and critical driving situations. The identification and analysis of corner cases and the targeted parameterization of test scenarios not only identify potential risks at an early stage, but also ensure the highest standards of vehicle safety. With these advanced methods, the automotive industry is ideally equipped to meet the challenges of the future and pave the way for safe, autonomous mobility.

 

imbus: Support for scenario-based testing and beyond

Regardless of whether it is a driverless vehicle or an autonomously operating robot - the people who come into contact with this system must be able to rely on the fact that such an autonomous system fulfills its tasks without errors and is safe.

imbus not only brings extensive expertise in scenario-based testing to the HolmeS3 project, but also decades of experience in software testing and software quality. As a leading provider in quality assurance and test automation, imbus supports companies in developing robust and reliable software solutions. With its own test management tool, the imbus TestBench, the execution of complex test scenarios is optimally controlled, from planning to automated execution.

imbus supports the introduction and integration of the HolmeS3 tool chain into existing development processes and also offers tailor-made solutions for customizing. In this way, imbus ensures that the implementation not only runs smoothly from a technical perspective, but also meets the highest quality standards.

Get advice now!

This might also be of interest to you: