OverFlowLight Prevents Gridlock, Optimizes Traffic Signals in Cities.

Urban traffic congestion often leads to severe queue overflow, where vehicle lines exceed intersection capacity, blocking upstream traffic and causing cascading gridlocks. Existing traffic signal control (TSC) algorithms, primarily focused on maximizing throughput, frequently fail to address these overflow situations during peak hours, exacerbating congestion and creating safety hazards. To combat this, a new real-time framework called OverFlowLight has been developed. OverFlowLight is designed to preemptively resolve overflow and enhance overall TSC performance. It utilizes multi-modal sensing from cameras and radars to accurately detect overflow in real-time. Upon detection, the system dynamically generates and inserts specific "overflow phases" into the signal cycle to clear the blocking queues. This is managed by a hybrid control design that combines rapid, rule-based overflow intervention with backend controllers, such as reinforcement learning, for optimizing longer-term efficiency. Extensive real-world deployments across 43 intersections in three major cities have shown that OverFlowLight seamlessly integrates with existing RL-based TSC agents, reducing overflow incidents by 60.4% and increasing network throughput by 18.2% compared to current baselines, while also significantly decreasing the need for manual intervention.