A scalable dynamic parking allocation framework

Cities suffer from high traffic congestion of which one of the main causes is the unorganized pursuit for available parking. Apart from traffic congestion, the blind search for a parking slot causes financial and environmental losses. We consider a general parking allocation scenario in which the GPS data of a set of vehicles, such as the current locations and destinations of the vehicles, are available to a central agency which will guide the vehicles toward a designated parking lot, instead of the entered destination. In its natural form, the parking allocation problem is dynamic, i.e., its input is continuously updated. Therefore, standard static allocation and assignment rules do not apply in this case. In this paper, we propose a framework capable of tackling these real-time updates. From a methodological point of view, solving the dynamic version of the parking allocation problem represents a quantum leap compared with solving the static version. We achieve this goal by solving a sequence of 0–1 programming models over the planning horizon, and we develop several parking policies. The proposed policies are empirically compared on real data gathered from three European cities: Belgrade, Luxembourg, and Lyon. The results show that our framework is scalable and can improve the quality of the allocation, in particular when parking capacities are low.