Research on the Choice of Strategy for Connecting Online Ride-Hailing to Rail Transit Based on GQL Algorithm

As traditional connection studies ignore the unbalanced distribution of connection demand and the variability of connection situations, this results in a poor match between passenger demand and connection mode, increasing passenger travel costs. Combining the economic efficiency of metro network operations with the unique accessibility advantages of ride-hailing services, this study clusters origin and destination points based on different travel needs and proposes four transfer strategies for integrating ride-hailing services with urban rail transit. Four nested strategies are developed based on the distance between the trip origin and the subway station’s service range. A reinforcement learning approach is employed to identify the optimal connection strategy by minimizing overall travel cost. The guided reinforcement learning principle is further introduced to accelerate convergence and enhance solution quality. Finally, this study takes the Fengtai area in Beijing as an example and deploys the Guided Q-Learning (GQL) algorithm based on extracting the hotspot passenger flow ODs and constructing the road network model in the area, searching for the optimal connecting modes and the shortest paths and carrying out the simulation validation of different travel modes. The results demonstrate that the GQL algorithm improves search performance by 25% compared to traditional Q-learning, reduces path length by 8%, and reduces minimum travel cost by 11%.