Collaborative passenger flow control for an urban rail transit network

Amid worsening passenger congestion, passenger flow control has become a major need in urban rail transit. However, existing passenger flow control strategies are fixed and do not consider the impacts of the spatial and temporal variations in passenger flow. To address the problems mentioned, a method for collaborative passenger flow control in urban rail transit is proposed. First, the temporal and spatial distribution of passengers in the network is obtained. Then, a model of collaborative passenger flow control for urban rail transit is built to reduce the number of controlled source points and controlled passengers and minimize the difference between the expected number of passengers on bottleneck links and the number of passengers with control. Next, a multilevel network that considers the time dimension is established based on the model. A forward-backward algorithm is introduced to make full use of the adaptive learning rate to solve the model. A case study based on a small-scale network shows that the forward-backward algorithm has good convergence. To verify the effectiveness of the method, the passenger flow of Chengdu Metro during COVID-19 is analyzed. The objective function of the forward-backward algorithm is 44% lower than that of the gradient descent method and 36% lower than that of adaptive moment estimate (ADAM). Its computational speed is also acceptable. The results show that compared with the fixed passenger flow control strategy of metro operators, the obtained strategy is more effective in reducing the number of controlled passengers and source points and alleviating congestion of bottleneck links.