Optimization of Traffic Signal Settings by Mixed-Integer Linear Programming

Remits obtained in Part I are extended to include offsets, splits at each intersection, and a common cycle time for the network as simultaneous decision variables. In addition to the deterministic link performance function, the stochastic effects of overflow queues from one cycle to the next are modeled by means of a saturation deterrence function that enters as an additive component in the objective function. Computational results demonstrate the feasibility of using mixed-integer linear programming on problems of realistic size. Sensitivity analysis of cycle time shows it to have a strong influence on network performance.