Development of optimization models for signalized intersections during oversaturated conditions

Abstract

This dissertation documents the development of optimization models for the control of signalized intersections during oversaturated traffic conditions. A dynamic optimization model was developed as the principal product of this research. The dynamic model's control objective is to provide maximum system productivity as well as minimum delay for a selected roadway system. A special feature of this model is its ability to manage queue lengths on external approaches up to predetermined upper limits. The model formulated to handle problems involving two intersections was extended to control three different roadway types: a conventional diamond interchange, a three-level diamond interchange, and an urban arterial with a critical intersection. The optimization model took the form of mixed integer linear programming. The control strategies generated by these modified dynamic models were compared with those derived from conventional signal timing models, using the TRAF-NETSIM microscopic simulation model. It was found that the modified dynamic models successfully produced optimal signal timing plans for the various oversaturated signalized intersections. The dynamic models consistently outperformed the conventional models with respect to system productivity. This conclusion was drawn from the TRAF-NETSIM simulation. The dynamic model solutions significantly reduced total system delay for most test cases, while slightly increasing the delay for a few test cases. Queue management on external approaches is a primary concern in the control of congested signalized diamond interchanges. The dynamic model was found to be superior to the conventional models in queue management for the congested interchanges. Queue spillback to upstream intersections is a major concern in the arterial systems having a critical intersection. The dynamic model solutions produced a significantly smaller queue spillback than the conventional model, TRANSYT-7F. The dynamic model controls queue lengths by the efficient and timely changing of signal timing plans as demand changes. The traffic control strategies presented in this research were also designed to minimize the transitional delay from frequent changes of the timing plans. The control strategies appeared effective in reducing this delay.