Development and testing of computational procedures for signal timing design at isolated intersections

Traffic control signals are an important element of urban street systems. While they are not, in themselves, the cause of delay and other problems perceived by the motorist, their operation has a very strong influence on the performance of these systems. There are several elements of the signal operation that may be addressed to improve the performance at any given location. One of the key elements is the accuracy of performance estimators themselves. Another is the manner in which the performance estimators are used to achieve the “best” signal-timing plan. Both of these elements are addressed in this dissertation.; The Highway Capacity Manual (HCM) describes a procedure for evaluating the performance of a signal in terms of control delay, defined as the amount of delay that would be eliminated for a given traffic movement if the signal control were eliminated from that movement. The HCM signalized intersection model represents the state of current knowledge of traffic signal operation. This dissertation explores the potential for advancement by addressing three specific shortcomings of the HCM model and its application: (1) the accuracy of the estimation procedure for the average green times for traffic actuated control, (2) the need for an iterative technique that can achieve the “best” performance of an intersection as evaluated by the HCM procedure and (3) potential improvements to the HCM model structure through the use of a “time-scan” technique to represent queue accumulation and departure. The following activities are described in the dissertation: (1) The functional and computational requirements for a formal optimization model were proposed, and an algorithm to meet those requirements was developed and tested. The optimization procedure used a hybrid genetic algorithm, involving a combination of a pure genetic algorithm and a “hill-climbing” approach. (2) The HCM procedure for average phase-length estimation was modified to improve its performance. Comparisons were made between the signal timings designed by the existing and proposed methods. (3) A time-scan procedure for representing the accumulation and discharge of queues on an approach to a signalized intersection was developed and tested using a hypothetical example involving permitted left turns.; The proposed enhancements were evaluated using the CORSIM simulation model as a surrogate for field data collection. The results indicated that the proposed optimization scheme was able to achieve its design objectives, the green time estimation procedures were improved by the proposed enhancements, and the accuracy of the delay estimates for approaches involving permitted left turns was improved by the proposed time-scan model structure.