.png){kind=icon}
.png){kind=icon}
Return
Ride-Hailing Assignment in Heterogeneous Networks Based on Graph Convolutional Neural Networks
.png){kind=icon}
PRE
AI
B
D
S
Z
A
DOI：10.1109/tits.2026.3692730.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The rapid growth of online ride hailing services has greatly improved passenger convenience. Existing methods that combine travel time prediction with order matching mainly focus on interactions between adjacent road segments, while ignoring latent relations between non-adjacent segments. In addition, global matching for mixed orders wastes computation on invalid and low-quality solutions. To address these issues, this paper proposes an online assignment framework for mixed ride hailing orders. First, a Graph Convolutional Neural Network with physical and virtual graphs is developed to extract heterogeneous road network features and predict travel time. Second, graph clustering and bipartite matching are combined to group and match mixed orders. Experiments on the urban road network within Beijing’s Fifth Ring Road show that, compared with baseline methods, the proposed method achieves higher travel time prediction accuracy and improves both the feasibility of matching results and online solving efficiency.
Keywords:
Graph neural networks
heterogeneous networks
intelligent transportation
ride-hailing assignment
travel time prediction
Journal
IF:
8.4
Papers:
9.5K
Citations:
6.3W
Organization
Cited Papers