Multi-criteria signal timing control strategies for critical intersections

Traffic Delay and queue are principal performance measures of effectiveness (MOE) to determinate level of service (LOS) at signalized intersections. In unsaturated conditions, delay minimization is appropriate for optimizing performance of signalized intersections. However, delays and queue sizes that individual vehicles experience at signalized intersection approaches are often subjected to highly stochastic and time-dependent variation. Previously the variability of MOE is not adequately studied. Also during over-saturated conditions, minimizing delay alone may not be effective since queue spillover may predominantly impose adverse effects on system performance. To develop effective and reliable signal timing strategies, reasonably accurate but tractable representations of these MOEs are required. Models for predicting the mean and variance of delay and queue are first presented in this thesis. The formations of analytical variability and approximate variability are developed for transportation researchers and engineers respectively. And the control strategies based on minimization of the variability are also formulated. In addition to delay minimization, strategies such as queue management, cycle failure, number of stops are considered as candidates for composing proper signal timing strategies under saturated conditions for isolated critical intersections in this thesis. Then comparison studies among these control strategies show that delay minimization and queue management are better strategies in all scenarios considered. The multi-criteria decision-making (MCDM) methods, including deviation minimization and compromise programming are adapted, to develop compromise signal control strategies and investigate the system performance of signalized intersections under various control criteria. Results of traffic simulation software CORSIM and developed Monte-Carlo simulation study indicated that these methods are capable of generating effective timing solutions fairly close to Pareto optimality.