Fixing traffic in Zurich: a data-driven control approach

Abstract

Perimeter control based on the Macroscopic Fundamental Diagram is a well established paradigm in the field of urban traffic control. The optimization problem aims at maximizing the flow in the city regions by bringing the densities near the critical point of the MFD. Several algorithms have been developed to solve this problem, most of them relying on model based predictive control. These approaches work well both in congested and uncongested settings, however they come at the cost of expensive modeling efforts especially when dealing with large scale systems such as a city. In order to solve this problem we apply a novel data driven algorithm (DeePC) to a real scale simulation of the city of Z¨urich, we test this algorithm against linear model predictive control (LMPC) obtaining better performance with no modeling effort.