Multiobjective and multimodal adaptive traffic light control on single junctions

Abstract

Most existing systems developed for traffic light control either consider a single objective or use a simple weighted sum to aggregate objectives. Three criteria are aggregated using an alternative way. The first two criteria refer to total delay dealing with two types of users with conflicting preferences: private vehicles and public transport vehicles. The third one (number of stops for private vehicles) aims at capturing pollution aspects. We propose a new mixed integer linear programming (MILP) formulation. Unlike most proposed MILP dealing with traffic light control problems, our main variables are traffic light switchover dates, during a time horizon. This permits a significant reduction of the number of variables. In order to take account of the traffic, we integrate this model within an interactive scheme, where information on the traffic is used periodically for adapting the control.