The field of traffic signal control is moving towards more adaptive and intelligent systems, leveraging advancements in reinforcement learning and graph-based models to optimize traffic flow and reduce congestion. Recent developments have focused on single-agent frameworks, which offer improved scalability and centralized management, as well as the integration of probe vehicle technology for more accurate traffic monitoring. Noteworthy papers include: A Dual Large Language Models Architecture with Herald Guided Prompts for Parallel Fine Grained Traffic Signal Control, which proposes a novel dual LLMs architecture for fine-grained traffic signal control, achieving a 20.03% reduction in average travel time. Designing Non-monetary Intersection Control Mechanisms for Efficient Selfish Routing, which presents a non-monetary mechanism for incentivizing socially efficient routing, demonstrating up to a 68% reduction in the efficiency gap between equilibrium and optimal flows.
Traffic Signal Control and Optimization
Sources
Graph-Attentive MAPPO for Dynamic Retail Pricing
A Dual Large Language Models Architecture with Herald Guided Prompts for Parallel Fine Grained Traffic Signal Control
Single-agent Reinforcement Learning Model for Regional Adaptive Traffic Signal Control
Robust Single-Agent Reinforcement Learning for Regional Traffic Signal Control Under Demand Fluctuations
Designing Non-monetary Intersection Control Mechanisms for Efficient Selfish Routing
Optimizing Multi-Lane Intersection Performance in Mixed Autonomy Environments