6 projects across simulation and big data.
Bi-directional green-wave optimization on an urban arterial using SUMO + TraCI; average delay reduced by 18.6% in simulation.
Processed 12M ride-hailing GPS traces to extract speed-density features and identify congestion propagation patterns via unsupervised clustering.
Trained a DQN signal controller in SUMO; delay reduced by 12% vs fixed timing under non-stationary demand.
Built a 32-node simulation network for an urban district, identified bottleneck links and proposed improvements.
Spatio-temporal clustering on 8M ride-hailing orders, identifying 6 typical travel patterns and tidal characteristics.
Built a road-network graph and used ST-GCN to forecast 15-min flow with MAPE 9.8%.