Robust optimization of intersection capacity

The optimal design of traffic light settings assumes in general that all the input parameters are known and constant for each design reference period, neglecting the effects of the uncertainty of the incoming flows or of other intersection characteristics, such as the saturation flows or the lost time. Responsive traffic plans only partially solve this problem, since they cannot vary continuously the traffic light settings, and generally consider only the input flow variations. The aim of this paper is to reformulate three well-known models for intersection capacity optimization in terms of stochastic programming and discuss their application to traffic light design of a real world intersection in Genoa.