To make the semi-synthetic data closer to the actual scenario, we can collect the routes of pedestrians from roadside surveillance cameras or the positioning systems on telephones. Because the routes of pedestrians involve privacy issues, we use an unmanned aerial vehicle to collect the pedestrians and vehicles data from two intersections in Gaoxin District, Zhengzhou, China. The routes of pedestrians in the semi-synthetic datasets are generated according to the data we collect, which conforms to pedestrians’ movement patterns and the ratio of pedestrians to vehicles in the real world. The routes of vehicles in the semi-synthetic datasets are generated by analyzing the records from roadside surveillance cameras in Jinan and Hangzhou, China. The more detailed information of the datasets is described as follows.
Jinan datasets: The datasets comprise the actual traffic flow in Dongfeng Sub-district, Jinan, China. The road network contains 12 intersections. There are 6295 vehicles entering the road network in one hour in Jinan datasets. Based on statistics information at the intersection with large traffic volume, we analyzed the ratio of pedestrians to vehicles to be 1:1. Therefore, we construct the routes of 6295 pedestrians.
Hangzhou datasets: The datasets comprise the traffic flow in Gudang Sub-district, Hangzhou, China. There are 16 intersections in the road network. The number of vehicles in Hangzhou datasets are much smaller than that in Jinan datasets, with only 2983 vehicles in one hour. Based on statistics information at the intersection with small traffic volume, we analyze the ratio of pedestrians to vehicles to be 1.325:1. Therefore, we construct the routes of 3952 pedestrians.