Videos for the Paper "Scalable evaluation methods for autonomous vehicles”

The field experiment was conducted at the intersection between Xuanhuai Avenue and East Nanyang Road on the campus of Shanghai Jiaotong University. In the real vehicle testing, there are three experimental vehicles, a bicycle, and a pedestrian, to participate in the experiment. In the first experiment, vehicles were tested in the initial naturalistic traffic scenario. The red box indicates the vehicle under test, the yellow box indicates the background vehicle, the blue box indicates the background bicycle, and the green box indicates the background pedestrian. The vehicle under test drives into the intersection area from the left side of the intersection, and then turns left through the intersection. Other background traffic participants enter the intersection area from the right side of the intersection and then go straight through the intersection. There is a trajectory conflict between the tested vehicle and other traffic participants near the center of the intersection. The videos below show the automated testing of real vehicles, including the automated detection and tracking of traffic participants around the tested vehicle, the acquisition of test data, the processing and analysis of test data, the automated calculation of test results and further output of test reports.

The tested vehicle was tested in a more complex scenario and was given a much lower score.

The tested vehicle was tested in the critical boundary driving scenario. There is a serious traffic conflict between the tested vehicle and the surrounding traffic participants in the video, but there is no collision. The tested vehicle was given a much lower score in the critical boundary driving scenario. By combining the data from the three real-world testings described above, the comprehensive evaluation scores and the final evaluation scores were given to the tested vehicle in the experiment.