Proper operation of traffic control signals requires signal timing plans that reflect current traffic and pedestrian needs. However, traffic signal timings are generally not reviewed as often as they should be and updated. It is estimated that improper traffic signal timings account for five to 10 percent of all traffic delay, on major United States arterials. The performance of signal systems highly depends on the methods and procedures used in the evaluation process. This research investigates and addresses limitations in current traffic signal timing evaluation methods and delay computation procedures.

In most cases, practitioners utilize analytical tools to evaluate traffic signal performance and develop signal timings that are implemented in the field. Performance of the optimized signal timings developed by any tool depends on the fidelity of the traffic models embedded in the tool. The quality of signal timings optimized by a variety of macroscopic and microscopic tools has not been investigated.

A fundamental step in updating signal timings is the estimation of the benefit of retiming, i.e., reduction in delay. This is the measure many agencies use to determine whether or not to update signal timings. Alternative methods that can quantitatively estimate the benefit of retiming signals are required, in lieu of the more complex analytical tools.

Control delay is the main metric for determining the performance of signal timings. The signal delay models in the 2000 Highway Capacity Manual (HCM) are inappropriate for complicated left turn phases such as permitted lefts and protected plus permitted lefts. There is a need for more accurate control delay estimation models, to improve traffic operations at signalized intersections.

The results indicate that, in overall, VISSIM-based Genetic Algorithm Optimization of Signal Timings (VISGAOST) and the Synchro programs produce signal timings of the highest quality and whose performance is very similar. The regression model developed has been found to reliably estimate the benefit of retiming signals. Results also show that, when the proposed modifications were applied to the IQA method, more accurate delay estimates were produced. Finally, more accurate incremental delay factor values have been derived using the new models developed in this research.