The transportation sector has been making great efforts in recent years to develop autonomous vehicles (AVs), which are envisioned to bring huge benefits in terms of safety, convenience, and precision. AVs rely heavily on sensors such as LiDARs, RADARs, and cameras to perceive the environment in order to navigate and identify objects. During adverse weather conditions, sensor performance degrades. However, the extent of degradation is not very well understood; one of the reasons is due to the lack of standardized testing methodology. Although outdoor testing is the most realistic, it faces issues, such as lack of control and repeatability, whereas indoor testing in a wind tunnel is more controlled and convenient to vary desired test conditions. A comprehensive methodology to quantify sensor perceptions in different weather conditions is presented. Traditionally, precipitation is simulated using spray nozzles, but this method tends to produce unrealistic droplet characteristics. Therefore, a more realistic precipitation simulation system is designed and implemented into both our full-scale and model-scale wind tunnels. The system is capable of simulating different droplet size distributions and the perceived intensities at the sensors during different driving speeds. Additionally, the methods of quantifying sensor visibilities and object detection accuracies inside a wind tunnel are discussed with example results, which are crucial elements for the operations of the AVs.
Session:
AERODYNAMICS
|
| 14:00 - 14:30