Design and optimisation of emergency bus shuttle services under urban rail transit disruptions

Public transportation can effectively mitigate the adverse impacts of rail transit disruptions on urban public transportation networks. This study proposes a multi-factor Long Short-Term Memory Networks (LSTM) prediction model integrating historical Origin-Destination (OD) data and weather conditions to forecast passenger demand during disruptions. It outperforms historical averages for high flows (>50 persons/hour). A hybrid optimisation model for emergency shuttle services combining full-stop, skip-stop, and direct-line modes is developed to minimise passenger time costs and operational expenses, considering constraints like transport capacity, dispatch frequency, and load factor. Taking the incident-affected section between Xidan and Wangfujing on Beijing Subway Line 1 as a case study, the implementation of a multi-mode emergency bus bridging strategy demonstrates a 6.8% reduction in integrated costs compared to conventional single-mode approaches. Additionally, a supplementary strategy using regular buses to handle partial passenger flow is proposed to optimise resource allocation when emergency bus resources are insufficient.