Integrated Dataset Generation and Testing Platform for Dynamic Graph Neural Network ETA Prediction

Dynamic GNNs need datasets that keep road topology, vehicle motion, and labels aligned. This paper presents an integrated platform: Traffic-DSTG-Gen builds route-aware dynamic graphs from SUMO, and TrafficLab evaluates ETA models in a real-time testing environment.

The generator emits graph snapshots with per-node and per-edge features for smart transportation research. The tester couples those graphs to interactive trips, model inference, and logged prediction error so models can be compared under controlled conditions.