🤖 AI Summary
This work addresses the challenges posed by heterogeneous road networks—such as diverse intersection layouts, varying signal coverage, and unknown topologies—by proposing a decoupled graph neural network (GNN) approach for traffic signal control. The method externalizes phase definition and timing logic from the learning module, employing a shared GNN to score traffic movements and mapping these scores to valid signal phases via a deterministic association matrix. This design decouples model parameters from both the action space of intersections and the graph size. Integrated with typed mean aggregation and proximal policy optimization (PPO), the proposed strategy demonstrates effectiveness on unseen synthetic grids and five real-world urban road networks, enabling transfer across heterogeneous networks, though it exhibits sensitivity under shifts in signal coverage distribution.
📝 Abstract
We present a traffic-signal control interface in which a shared graph neural network assigns scores to individual traffic movements. Each junction converts these scores into its own variable-sized set of legal signal phases using a deterministic incidence matrix. Directed corridor nodes provide traffic context, while movement nodes represent controlled input-to-output paths through junctions. Typed mean aggregation produces one scalar per movement; phase definitions and signal timing remain outside the learned network. This makes graph size and junction-specific action count independent of the learned parameter shapes. PPO experiments evaluate the interface on unseen synthetic grid geometries, altered signal coverage, and five heterogeneous city graphs. The policies retained performance across unseen geometries within the synthetic grid family, while changes in signal coverage exposed sensitivity to a signal-coverage distribution shift. A single trained city-policy instance executed across all five city graphs, with heterogeneous outcomes. These results provide feasibility evidence rather than a general estimate of transfer to arbitrary road networks.