Modeling the Impact of Passenger Population Heterogeneity and In-Vehicle Congestion on the Stochastic Passenger Flow of Metro Networks

The passenger flow pattern serves as crucial information for the planning and operation management of metro networks. Nevertheless, the passenger flow derived from the existing passenger flow assignment models is deterministic and overlooks the stochasticity of passenger flow resulting from different individual path choices among heterogenous passenger population. This study examines how stochastic entry, exit, and transfer walking times, which are caused by the different walking speeds of heterogeneous passenger population, affect the stochastic aggregate prediction of path choice and the equilibrium probability distribution of passenger flow. An aggregate prediction of the path choice model considering stochastic path travel time and a passenger flow equilibrium assignment model considering aggregate prediction among heterogeneous passenger population are provided in this paper. Then, a method of successive averages based on Monte Carlo simulation is presented to produce stochastic equilibrium passenger flow. Moreover, a third-order polynomial normal transformation based on L-moments is introduced to universally characterize the complete percentile distribution of path or link passenger flow in a network. The results of case studies including the Shanghai Metro network indicate that: (1) the stochastic walking times caused by heterogeneous passenger population indeed have an impact on the aggregate prediction of path choice; (2) the greater the stochasticity in walking time, the greater the effect on path flow stochasticity; and (3) stochastic walking times affect the stochastic passenger flow in both transfer links and in-vehicle links. The proposed model accounts for the aggregate prediction of path choice among heterogenous passenger population in passenger flow predictions. The derived stochastic passenger flow characteristics enable the upgrading from conventional point prediction of passenger flow to confidence interval prediction.