VW to use virtual validation

In the future Volkswagen intends to make new driver assistance systems production-ready using virtual validation.

The next generation of assistance systems will then learn from virtually generated driving and traffic situations. The brand expects this approach to make development processes even faster and more efficient.

In the long term, it is conceivable that millions of test kilometres required for validating automated driving could be completed in virtual environments. Experts from Volkswagen are already testing software developed in- house to simulate driving in traffic.

‘We are continually developing Volkswagen vehicles and taking innovations into all segments,’ said board member for development Dr Frank Welsch. ‘We are building on our strong global development team and grasping all the opportunities offered by digitalization. This also includes virtual validation. We are developing this technology for our work as it will make for faster and more efficient development processes.’

Volkswagen is aiming for two main benefits with virtual validation. Firstly, assistance systems can be trained continuously over days and weeks in any scenario. This approach dramatically accelerates the learning speed of the systems concerned.

With virtual validation, Volkswagen also expects to be able to develop a rapidly growing number of systems and networked vehicle functions to production maturity.

To date, assistance systems have been tested using a hardware-based approach by connecting components to test rigs via data interfaces. As the number of networked functions grows, this means more and more hardware-based tests are necessary. Virtual validation will reduce the volume required, as physical test rigs will no longer be essential.

A new assistance system will then be connected to these virtual scenarios. Its sensors will process the virtual ambient data in the same way as actual ambient conditions. The software will also visualize the virtual scenario via a 3-D graphic environment. The engineers will then be able to observe the behaviour of assistance systems precisely and to intervene and optimize the systems as required.

Volkswagen is already testing the software which has been developed in- house. The first application simulates thousands of individual car parks with freely definable parameters (architecture, lane guidance, traffic, etc.). Car parks are regarded as an ideal example of complex environments which must be mastered by an assistance system.

In the long term, it is conceivable that millions of test kilometres required for validating automated driving could be completed in virtual environments. The self-learning systems of the vehicle (artificial intelligence) would process this data in the same way as data from physical tests on proving grounds and public roads. This would further accelerate the development of production-ready automated driving functions.