Lane-formation in counter-flow based on DBSCAN

With 1 the development of society, more and more large-scale activities are conducted in the cities. The issue that how to avoid emergency in respect of human performance has attracted a lot of researchers. It is easy to find the counter-flow in large-scale activities. Bi-directional movement has attracted a lot of researchers to investigate. More attention is paid to the existence of lane formation in respect of experiments and simulations. However, few attentions are drawn to the definition of lane formation in counter-flow. In this paper, a method based on deep learning technique(DBSCAN) is proposed to define lane formation. Lane formation based on DBSCAN is used to gather the pedestrians in a group, and order parameters and time step are taken into consideration in order to obtain the number of lane formation explicitly. Then proposed method is compared with previous lane formation definition. All the lanes defined by previous method are quantified during movement process, which could reflect self-organization in time, while method introduced in this manuscript are measured after movement, which needs more data after event. The proposed method has the ability to calculate the number of the lanes. The work in this paper is intended to better understand movement during counter-flow in a corridor and develop a technical foundation for evacuation strategies in large-scale activities.