Xylon is demonstrating its all-in-one fourth-generation datalogger, Quattro, at Automotive Testing Expo North America. The setup showcases acquisition from 14 x 2.3-8MP video sources, including cameras with FPD Link-III and GMSL2 automotive interfaces, GigE Vision and USB reference cameras, as well as synthetic video channels streamed from the simulator. This configuration uses less than 20% of the logger’s data bandwidth and interfacing capacity, and also includes the car’s ethernet, CAN and LIN networks.
Xylon’s CEO, Davor Kovačec, noted that development and validation of AI-based automotive ADAS/AD systems require huge volumes of data acquired directly from a vehicle’s heterogeneous sensors or from simulation environments. For use in machine learning and AI validation, the acquired data must be carefully distilled into training and golden truth data sets.
Data processing – from acquisition to AI implementation – involves many challenges. Road data acquisition from realistic vehicle installations, in which the AI system interfaces with dozens of sensors, must be non-invasive, Kovačec explained. Terabytes of data from video cameras, radars, lidars and simpler sensors must be centrally timestamped and recorded with preserved phase alignments. Post-processing extraction of interesting sequences and build-up of coherent data sets must be quick and cost-saving. And finally, distilled data must be stored in industry-standard formats for simple use in third-party platforms for machine learning, test and validation.
“Our experience in automotive testing and validation is organically expanded with our R&D in AI, programmable chips and embedded systems,” said Kovačec. “Not only has this enabled us to develop a multifunctional and ultimately configurable platform for precious data processing but it has also enabled us to better understand our clients’ needs and help them achieve specific goals in new and innovative ways.”
In terms of synergies with other exhibitors, Xylon works with third-party providers of simulation platforms with a common goal of immersing real hardware ECUs into synthetic simulation worlds. In such advanced HIL systems, Xylon dataloggers work as an interface between simulators and ECUs in fully closed-loop virtual test drives.
To see a demo of Quattro, visit Booth 1004.
Click here to register for your free expo pass.