ZF Revolutionizes Driver Assistance Systems with AI

As one of the leading AI solutions, “ZF Annotate” has been in use for more than three years and is attracting growing interest from automakers. How does the technology help to make measurement data more usable – and, in doing so, enable driver assistance and autonomous driving functions to be developed more efficiently and faster? 

Our technology helps automakers make measurement data far more usable and comprehensive. The AI automatically generates a consistent, highly precise environmental model from raw data. All relevant objects – such as vehicles, pedestrians, traffic signs, or lane markings – are reliably detected, classified, and tracked over time. As a result, processes that previously had to be carried out manually and were only limitedly scalable become reproducible, stable, and consistently high in quality. The cloud-based architecture also makes it possible to efficiently analyze large volumes of sensor data collected from global test fleets covering distances of more than hundreds of thousands of kilometers. Within a short time, this produces actionable reference data from test drives totaling up to 4,000 kilometers per day. Manufacturers can thus shorten development cycles, generate training data, and accelerate the data-driven development of driver assistance and autonomous systems.

What measurable results does the AI solution deliver in concrete terms, particularly with regard to data accuracy, efficiency, and speed? 

“ZF Annotate” delivers verifiable improvements across key development parameters. These include more robust sensor calibration and a unified time base for synchronization. Through AI-based object annotation, classification with additional required attributes, such as size and speed, and 2D and 3D tracking algorithms, the system generates a highly precise “ground truth”. In practical terms, this means that all relevant objects are captured with positional accuracy and tracked with stable IDs across entire driving scenes. Previously, this level of precision could only be achieved through substantial manual effort. In terms of speed, AI accelerates the validation process by a factor of one to up to ten. Comprehensive annotation projects, for example, can be shortened from twelve months to just two. At the same time, cloud-based processing of large data volumes significantly increases efficiency. The fully automated analysis reduces the effort required to create reference data by up to 80 percent, enabling rapid delivery of reliable reference datasets for development and validation teams.