Smart Solutions for ADAS Development
NESOSA has been developing Advanced Driver Assistance Systems (ADAS) for years and has successfully launched these innovations into the market. Now, the technology company offers the same digital tools and services it tested directly to its customers, accelerating ADAS system development and reducing costs.
Recognizing that a mailbox is not a pedestrian and that a parked car doesn’t move is second nature for humans. This basic understanding is crucial for safe and comfortable navigation in traffic. Advanced Driver Assistance Systems perform similar tasks, correctly recognizing and processing objects such as other road users, lanes or traffic signs in real time. Cameras, radar, lidar or ultrasound sensors continuously provide data from which the vehicle generates a three-dimensional image of its surroundings. It is important that the vehicle processes the digital sensor data correctly so that the vehicle obtains the “ground truth.” If this “ground truth” is not true, it can lead to serious consequences.
Ensuring "Ground Truth" with Artificial Intelligence
How can we ensure that the system processes the data correctly and that the “ground truth” actually corresponds to reality? During the development and validation process, reference data is collected to compare with the results of the systems being tested. For this purpose, ZF has developed the ZF Annotate validation service. The reference measurement is performed with an additional sensor set that collects data independently of the setup to be tested.
The sensor set can either be mounted on the test vehicle itself or on a separate vehicle driving behind it and is confronted with the same traffic situations. The recorded data is then uploaded to the cloud and analyzed. Thanks to artificial intelligence, all relevant objects are accurately marked, classified, attributed and assigned unique ID numbers. Moving objects are tracked. With this annotation, ZF Annotate offers a highly precise comparative measurement.