PROJECT NUMBER: PSR-26-15

FUNDING SOURCE:

TOTAL COST: $110,000

PERFORMANCE PERIOD: SEPTEMBER 2026 - SEPTEMBER 2027

Project description

Traffic bottlenecks are among the primary contributors to travel delay, unreliable freight movement, and elevated crash risk in urban transportation systems. The goal of this project is to develop a proactive, AI-enabled operational decision-support system that empowers transportation agencies to anticipate, mitigate, and evaluate emerging bottlenecks before they degrade mobility, freight reliability, and safety. The system will integrate both historical and real-time traffic information, combining high-resolution vehicle trajectory data with fixed-sensor observations from Caltrans’ Performance Measurement System (PeMS). Using these complementary data sources, the project will develop machine learning models that continuously monitor evolving traffic conditions and identify the onset of major bottlenecks, defined as congestion events that persist for multiple observation intervals rather than dissipating quickly. The output of the early warning module will be a real-time classification of bottleneck severity and persistence, which can provide traffic operators with advance notice of locations where significant congestion is likely to develop. Building upon this predictive capability, the project will conduct a pilot deployment that demonstrates how early warnings can support proactive traffic operations. The pilot will assess the usability, timeliness, and perceived value of the early warning capability through collaboration with agency partners, generating practical guidance for integrating artificial intelligence (AI) and predictive analytics into routine traffic management practice.