Development of a Traffic Signal Green Time and Cycle Length Optimization Model Using NSGA-III

Abstract

Optimization of delay, capacity, and emissions in signalized intersections brings in conflicting goals. But most literature studies have dealt with these individual goals or used two-objective optimization methods, and such approaches are not efficient in capturing these tradeoffs. To overcome this problem, this research utilizes Non-dominated Sorting Genetic Algorithm-III in designing a model to optimize delay, capacity, and emissions for a four-leg intersection. Convergence, hypervolume indicator, and spread methods are used to examine algorithm performance, and knee point solutions are used to identify a tradeoff solution. The obtained output shows that NSGA-III gives a smooth and evenly spread Pareto front with a hypervolume  and a spread of  which represents excellent convergence and diversification capabilities. Following the outcome of the experiment based on knee point identification, the solution with index 38 gives an optimal control setting of q₁ = 37.7 s, q₂ = 20 s, q₃ = 20 s, q₄ = 27.8 s, with cycle length of C = 121.42 s. With this setting, the average delay is reduced by 38%, the queue length by , and the degree of saturation is improved to 0.83, while capacity is reduced moderately and total emissions increased. In summary, this research work proves the NSGA III is efficient in identifying tradeoffs among delays, capacity and emission. The highlight of this research work is that the knee point gives the most balanced operational solution without excessively increasing cycle length.