ABSTRACT. Electric buses play a critical role in advancing urban green transportation. However, a transition period remains before the full conversion to electric buses is achieved. During this period, it is essential to address the scheduling challenges associated with operating a mixed bus fleet of electric and diesel buses. This study considers the scheduling problem for mixed bus fleets using a time-space network modeling approach. To evaluate the proposed model, test instances of varying sizes are generated based on real-world data. The findings provide practical insights and serve as a reference for bus companies managing mixed fleets in their scheduling operations.

ABSTRACT. Bike-sharing has emerged as a crucial first- and last-mile solution in urban mobility, particularly around subway stations where multimodal connectivity is essential. Accurate demand forecasting and intelligent dispatching of shared bikes in these areas present significant challenges due to complex spatial-temporal dependencies. To address this issue, this study proposes a collaborative prediction model—CoGLSTM—for bike-sharing demand, considering its spatiotemporal dependence on subway travel demand. Specifically, metro passenger flow, bike-sharing trip orders, and built environment data are fused to analyze usage patterns of bike-sharing around metro stations. A Graph Convolutional Network (GCN) is employed to model spatial heterogeneity among stations, while a Long Short-Term Memory (LSTM) network captures temporal evolution. Experimental results based on real-world data from multiple bike-sharing platforms in Shanghai show that the proposed CoGLSTM model substantially outperforms baseline approaches in short-term demand prediction. The model offers robust support for dynamic shared bike redistribution and operational optimization in urban transportation systems.