Reading Time: 4 mins



Jann-Eve Stavesand


Dr. Manaswini Rath, KPIT;
Claus Hellberg, MAN Truck and Bus SE;
Christer Neimock, T-Systems;
Hans Jurgen Holburg, BTC Embedded Systems AG;
Dr. Simon Robner, TUV SUD;
Tino Schulze, dSpace GmbH.

The validation of product safety especially with regard to autonomous systems – vehicles — is a huge challenge given the numerous situations the vehicle may find itself in. The panel discusses key questions — What are these challenges? How do different companies resolve these? How do they interact with each other to offer solutions for these?

For the OEMs, the central issue revolves around how to homologate. It’s clear why. Initially, their focus was on the vehicle-side so, these would be selected from the portfolio, then, discussed with the authorities and finally, tested on the test tracks before they were approved. Legislation was as critical as simulation so tests would be measured on efficiency maps and certified. Data would simulate in a vector tool developed by the EU. Since the tool was the same for all OEMs, the problem arose when certification of autonomous functions and vehicles was required.

The question always remained: How could certification and approval process be consistent across the board? How could adaptive functionalities be brought swiftly to vehicles?

From the tech perspective, the matter is complex. While regulations have been updated for highly automated driving yet, different approaches are adopted to ensure product safety — investing in safe systems that decrease effort needed to check complex, highly automated driving systems, being one. The question of scale also looms large. However, the “coverage” that tech service provides in testing the range of functionality has been seen to decrease when audits, processes and checks consistently measure product safety. This is where partnerships can help.

Adding value here are the software engineering solution providers like KPIT– critical cogs in the homologation and virtual simulation wheel — who bring scale in terms of infrastructure, scenarios (coverage of the corner cases, safety cases, road and traffic conditions etc.) and engineers with domain experience. Such providers validate autonomous systems well because they build these, have domain knowledge and the ability to create the strategy to develop those scenarios for the testing and test cases. Plus, as integrators, competence in virtual simulation and validation, scenario creation, testing, report generation, deep safety analysis and collaborating with tier ones who are the sensor developers, helps build a certain trustworthiness around the whole system validation process.

Other important aspects favouring collaboration as a step toward homologation are:

Building the pillars for overall product safety: Though the standards are constantly evolving, know what scenarios must be checked and use the freedom to go beyond this — to simulation, audit and testing.

Enable continuous testing: Function development may be quite disparate, but from the standpoint of homologation, a seamless transition of tooling — know where to get the device and test from – is crucial. That’s where Continuous Testing can help.

Assess coverage: This focuses on the testing – requirements-based and scenario- based. While the combinations of scenarios make it tough to define the coverage criteria for homologation, BTC, dSpace and MAN have introduced a paradigm shift in defining scenarios – the Abstract Scenario Definition — where a graphical language which is open scenario compatible is developed. This allows customers to define a set of abstract scenarios to be considered and covers the direction of homologation finally, based on a given ODD.

Leverage the Two-stage technology: This creates scenarios automatically from those abstract scenarios which can be defined graphically.

Stress the system under test: This will test the algorithm in the car — when undertaking virtual or real simulation. AI technology then, can observe the simulation and highlight the weakness function. Also, it can point out if the scenarios were simulated in the way specified in the abstract scenario. For example, if a reactive traffic controller that comes together with a test case is generated and in the dSpace environment, it proves its capabilities, the final solution may enable homologation.

Build the data software flow path: When improved and fully automated in terms of a DevOps process to fulfill the requirements, homologation and Verification- Validation will not take longer than 2 hrs. This is where software infrastructure providers and system integrators can provide and fulfill data flows on a scalable, continuous and fully automated platform.

There are challenges galore in the form of anomalies or when bringing this data into the cars and ingesting the process into the backends, which may require fast running of multiple data lakes that combine all aspects – training, validation, transport, retraining – in real-time in future.

But, collaboration between all stakeholders can enable implementation of innovative ideas to manage these challenges, work around the smaller residual risks and understand the bigger picture. Homologation is complex And partnerships are a key where players like infrastructure providers, engineering solution providers and tools companies come together and work along with OEMs and Tier1s to solve this challenge.

To know more about Safety Validation – The Path to Homologation, watch the panel discussion at

Copy link


Leave a Reply

Your email address will not be published. Required fields are marked *