AI project sharpens the senses of automated test vehicles

On 28 November, Bosch Research together with partners presented the results of the KIsSME project at the Bosch Research campus in Renningen. The acronym stands for: “Artificial intelligence for selective near-real-time recording of scenario and maneuver data in the testing of highly automated vehicles”.

The joint project was funded by the Federal Ministry of Economics and Technology and ran for three years. It dealt with a topic that is central to the development of automated vehicles: AI-based intelligent identification of critical traffic situations and the associated selective storage of data generated during test drives. Among other things, these data will be used to train maneuver algorithms in order to react even better to complex traffic situations in autonomous mixed traffic.

Project speeds up development of safe automated vehicles

In order to develop automated vehicles, test drives must take place during each phase. During these tests, numerous sensors record information about the environment and the status of the car. Enormous amounts of data — several gigabytes per second — are stored – more than can be processed using conventional processing and analysis methods. Part of the data is essential however in order to check and improve on the decisions made by the on-board systems during the journey.

As part of the KIsSME project, a software prototype was therefore developed that only stores the data from relevant and critical traffic situations that represent added value for the further development process. With the help of AI-based algorithms, the system fully evaluates each traffic situation in real time and decides whether and, if so, what information about it should be retained. Another special feature: The prototype can be adapted flexibly to meet the needs of developers. This allows the correct data to be recorded in test drives during the various stages of development.