Predicting Free Parking Slots via Deep Learning in Short-Mid Terms Explaining Temporal Impact of Features

Abstract

Looking for available parking slots has become a serious issue in urban mobility, since it influences traffic and emissions. This paper presents a set of metrics and techniques to predict the number of available parking slots in off-street parking facilities. This study deals with deep learning model solutions according with a mid-term prediction of 24 hours, every 15 minutes. Such a mid-term prediction can be useful for citizens who need to plan a car transfer well in advance and to reduce as much as possible any computational effort. Since most solutions in literature are focused on 1-hour ahead prediction, the proposed solution has been also tested in these conditions. The proposed solution is based on Convolutional Bidirectional LSTM models. Results have been compared in terms of precision metrics based both on occupancy and free slots. The paper also provides a framework to pass from an assessment model based on occupancy to models based on free slots and vice-versa. The obtained results have improved those already available in literature. A formal study has been conducted to perform feature relevance analysis by using explainable AI technique based on gradient and integrated gradient and proposing new heatmaps which highlighted the difference from LSTM and Bidirectional LSTM, feature relevance (base line, weather, traffic, etc.) and the impact of seasonality on predictions, namely the temporal relevance of features. The comparison has been performed on the basis of data collected in garages in the area of Florence, Tuscany, Italy by using Snap4city platform and infrastructure.INDEX TERMS smart city, available parking lots, prediction model, machine learning, deep learning, explainable AI.

Data Sources and Models for Integrated Mobility and Transport Solutions

“…are at the basis of any assessment of traffic conditions [ 15 ], such as counting the number of vehicles traveling on the roads (and distinguishing among the different kinds of vehicle), estimating traffic flow, identifying critical conditions and in some cases recognizing car license plates and thus verifying authorization, insurance, computing origin–destination matrices, etc. Other kinds of data are those related to parking sensors, which may communicate slot occupancy [ 16 ], and in some cases also may perform a match with the specific car park slot authorization. In most cases, automatic gates with cameras are used to count entities crossing specific areas, verify plates and authorizations and issue sanctions/taxations for entrance in restricted traffic zones and/or entering the city (for example for tourist buses).…”

“…Such processes primarily involve data-transformation tools to extract from basic data sources a set of more usable and valuable data models, which are actually more usable for deeper analysis. The operative management processes in the blue blocks include, for example, estimations of the data (listed in pink blocks) such as: KPIs (key performance indicators) such as Sustainable Urban Mobility Indicators (SUMI) [ 119 ] and the SUMP, Sustainable Urban Mobility Plan [ 120 ], required to assess city mobility and transport management conditions/facilities; Predictions of traffic flow [ 121 ], parking lots status [ 16 ], sharing service conditions, etc., which are typically produced by some deep learning models; Anomaly detections: for example, comparing real-time conditions with respect to typical or predicted conditions and thus producing notifications, tickets for maintenance and alarms when critical conditions/events are detected; Routing, multimodal routing and conditional routing for producing routing paths by taking into account real-time traffic/environmental conditions or possible changes inside city structures due to last-minute ordinance, accidents and natural/non-natural events; Origin–destination matrices (from census data, from OBU devices, from mobile apps data, from mobile operators’ data, etc., or by data fusion): trajectories for people and vehicles, semaphores cycles and simulations, in general; Prescriptions to solve critical conditions, such as improved semaphore cycles to reduce time to across the city, changes within city viability, etc. They are typically produced by using operative research algorithms exploiting optimization models.…”

“…Predictions of traffic flow [ 121 ], parking lots status [ 16 ], sharing service conditions, etc., which are typically produced by some deep learning models;…”