Improving delivery conditions by dynamically managing the urban parking system: Parking availability prediction

An adequate dimensioning of delivery bays combined with their effective availability facilitates deliverers' tasks, fluidizes urban traffic and improves living conditions in the city. Despite this marked interest, there is still no formal tool associated with a global approach to design and manage delivery areas network. In this paper, we lay the groundwork for a new vision of urban parking. It consists in sharing, equitably between the different road users, all the parking infrastructures. It considers that parking spaces can switch, in real time, their profile from a classic parking place to a delivery bay and vice versa. It is based mainly on the parking availability prediction. We employ three types of ensemble regressors to forecast the number of free parking slots. We examine their outcomes, evaluate their performance and determine their relative advantages as well as disadvantages. We conduct these tests leveraging parking data collected and published by the City of Melbourne, Australia.