With its high infection rate, COVID-19 has swept the globe and brought great challenges to social life and economies. As an essential form of public transportation, the Beijing subway plays an important role in transportation systems. In traditional subway organizations, all one-sided doors of a train carriage are employed for passengers’ alighting and boarding. A higher risk of COVID-19 infections may be attributed to inevitable bidirectional conflicts at doors with higher passenger volumes. Moreover, quantitative analyses for this problem and corresponding solutions are, limited in recent studies. In this research, conflicts at carriage doors are analyzed using a cellular automaton (CA) based model. Four schemes to separate alighting and passenger boarding into separate doors are investigated. The performances of different schemes with various alighting and boarding passenger ratios are simulated with the software package Legion Studio. Both macroscopic and microscopic parameters to characterize passenger conflicts are obtained for analysis. The separation of alighting and boarding passenger flows yields the desired reduction in bidirectional conflicts, which further limits the probability of infectious disease exposure. This is an important reference to improve current practices and provide specific measurements of passenger organization under abnormal situations.