An algorithm to minimize queues at signalized intersections

Previous research regarding optimization of isolated intersections has been carried out, yet optimization techniques implemented on the signal controller's level are unsatisfying. Three major obstacles prevent the controller from adapting programs to varying traffic conditions: (1) the necessity to deal with several constraints, such as inter-green times, minimum green times, stage (phase) sequence, green-wave constraints, etc.; (2) complex non-polynomial optimization problems, especially for saturated and oversaturated conditions; and (3) the difficulty in collecting the data required to control an intersection for oversaturated conditions. In this paper, a self-tuning system for controlling signalized intersections in undersaturated and oversaturated conditions is depicted. The system was successfully tested in one of Haifa's (in Israel) most congested intersections. The system tailors some well known control techniques, together with some specially developed methods and algorithms, to overcome the difficulties.