DATALOGGING No bits left behind Current autonomous driving system architectures require a whole new bag of tricks from datalogging equipment By Gordan Galic , Xylon No need for external converters – the Xylon Quattro directly connects to more than 60 automotive sensors, from high-speed video cameras and lidar to low-speed ultrasound and GPS sensors. Users can select IO modules with automotive interfaces to configure the datalogger for an exact application BOOTH: 6310 I f you think that the vehicle test datalogging challenge has been permanently solved, you might want to think again. Datalogging is becoming more complex with the introduction of new autonomous driving system architectures. Although some OEMs opt for a fully visual approach to understanding the environment surrounding the vehicle, the vast majority of Tier 1s/OEMs rely on heterogeneous sensor suites of camera/ultrasonic/radar/lidar. To fully comprehend the complexity of modern datalogging, let us dive into an illustrative example of a sensor suite and crunch some numbers. Imagine that the suite has eight 8MP video cameras for an improved driving experience, safety functions and autonomous navigation. Their high video resolution enables monitoring of the complete space surrounding the vehicle at different distances. Four additional 3MP surround-view video cameras enable easy parking. Lidar and radar sensors enable people and object sensing even in complete darkness and fog. ADAS/AD systems traditionally use fewer lidar sensors than cameras. The illustrated suite uses six lidar and four radar sensors. Besides high-bandwidth automotive sensors, the sensor suite includes common, simpler sensors such as 10 ultrasound sensors, GPS and digital and analog IOs. Unlike production vehicles, test vehicles use additional sensors for research and datalogging control purposes. The Below: Estimated data bandwidth needed for reliable logging of all sensors in the illustrative sensor suite suite from the example uses four reference video cameras to monitor the vehicle’s in-cabin area and specific exterior details such as road lanes. Datalogging equipment suitable for this specific sensory suite must fulfill requirements roughly divided into three groups: interfacing, data bandwidth and storage capabilities. Only top-level dataloggers such as the Xylon Quattro can reliably collect all precious data from the sensor suite at this level of complexity. The Xylon Quattro is a modular system and can be configured with different IO modules for native automotive interfaces. Four video slots accept four-channel video modules for up to 16 GMSL2, FPD-Link III and similar cameras. Lidar sensors and surveillance cameras can connect via four four-channel ethernet modules; radar and low-speed sensors connect directly via more than 40 CAN, LIN, UART and other automotive interfaces. Having all sensors directly connected enables super-precise timestamping and preserved data phase. Unlike the Xylon Quattro, ethernet-based dataloggers require several converters, such as GMSL2 to 10GigE, which significantly increase costs and system complexity. SENSOR TYPE 8MP camera 3MP camera 5MP camera Lidar Radar Other INTERFACE GMSL2 GMSL2 Ethernet; MJPEG Automotive ethernet CAN FD CAN, LIN, dig… NUMBER 8 4 4 4 4 20 SPECIFICATIONS 8MP 30fps 3MP 30fps 5MP 30fps 1.4 million points NA NA DATA BANDWIDTH 8x ~3Gbps = 24Gbps 4x ~1Gbps = 4Gbps 4x ~1Gbps = 4Gbps 6x 0.15Gbps = 0.9Gbps 4x 0.01Gbps = 0.04Gbps ~ 0.3Gbps TOTAL ~33Gbps The table shows estimated data bandwidth needed for reliable logging of all sensors in the suite. Most of the data is generated by video cameras and the total bandwidth is around a demanding 33Gbps. Not many datalogging systems can fulfill this requirement, but the Xylon Quattro can easily support data bandwidth two or even three times higher than that. The illustrated sensor suite generates about 4GB/sec, or about 14TB of data per hour. Consequently, the integrated storage system must store at least 100TB of data per day. With its four trays for exchangeable NVMe SSD disks, the Xylon Quattro offers enough storage for days of datalogging – up to 128TB. The example shows that even a smaller test fleet can easily record over 1,000TB/day. Post-processing of such data amounts requires significant resources that can be reduced by using a smarter approach. The Xylon Quattro implements sophisticated triggering that detects interesting events, and filters only interesting data. For example, it can be set to record valuable data 15 seconds before and 60 seconds after a speed limit sign detected by a forward-looking camera. This illustrative use case demonstrates how the right datalogging choice makes an impact beyond simple data harvesting and helps organizations through different development and testing stages. To find out more, scan the QR code or visit: www.xylon-lab.com Xylon 70 ADAS & Autonomous Vehicle International April 2024