ADAS.ai is based on two achievements: First, on an extremely high-resolution, multi-sensor synchronised dataset based on real-world driving of all scenarios and the corresponding number of kilometres needed to globally validate an L2+ ADAS system. And secondly, on a proprietary AI technology developed with ZF cooperation partner Cognata (Rehovot, Israel). This technology makes it possible to virtually "exchange" sensors, as the algorithms can convert the already recorded sensor signals within the high-resolution data set accordingly and thus simulate other settings. This eliminates the need for further test drives with real vehicles, as sensors can be tested and validated in virtual space.
ADAS.ai is cloud-based and offers significant cost and quality advantages compared to traditional development and validation based on physical test drives and existing simulation technology. Full L2+ scenarios derived from real-world driving are fed into the development and validation phases at multiple points - either in "virtual" or "full sensor" mode. This enables immense acceleration and efficiency of the development and validation processes. Incidentally, the service can be applied not only to ZF's own ADAS, but also to systems from other Tier 1 suppliers.
"We use real-world data in an end-to-end approach because we know what our customers and the authorities want," says Christophe Marnat, Executive Vice President of ZF's Electronics and Advanced Driver Assist Systems Division. "With our approach, an OEM does not need a large number of test drivers or millions of dollars in budget to develop and validate in a safe framework. We bring the real and virtual worlds together because we use real data to drive both modalities." Marnat: "Our vision is to enable fully virtual engineering for ADAS from all Tier 1 suppliers. OEMs want ADAS attributes to be as reliable as a crash test - fully digitally validated before the first prototype is built, and ZF has the technology to make that happen." ZF's ADAS Digital