The paper introduces HiLLTS, a zero-shot traffic signal control framework built around a three-level hierarchy: a central coordination agent, district-level control, and cluster-level intersection agents. The system is designed to transfer across road networks and demand regimes without extensive reinforcement-learning retraining. According to the abstract, compared with the strongest non-LLM baseline in each scenario, HiLLTS lowers average waiting time by 36.73% in low congestion and 14.71% in high congestion, while reducing CO2 emissions by 7.87% and 8.57%, respectively. Larger gains are reported against Fixed-Time and Max Pressure baselines.
No heat snapshots are available in the last 24 hours.