A digital twin framework for urban parking management and mobility forecasting.

Abstract

Rapid urbanization and population growth have created significant challenges in urban mobility management, such as traffic congestion, inefficient public transportation, and environmental pollution. Here we present the development and implementation of a digital framework for urban parking management and mobility forecasting. The framework integrates a wide range of historical and real-time data, including parking meter transactions, revenue records, street occupancy rates, parking violations, and sensor-based parking slot utilization. Additionally, the data encompass weather conditions, temporal patterns (such as weekdays and peak hours), and agent shift schedules. Descriptive statistics are used to identify key patterns, while the Spatial-Temporal Identity model is used for the predictive phase, and the Conditional Variational Generative Adversarial Network is used for the generative phase of the digital twin. The algorithm allows forecasting parking demand and mapping data for spatial planning and resource allocation. Moreover, the integration of the Generative Artificial Intelligence model generates realistic what-if scenarios for virtual testing of mobility strategies before real-world implementation. The results highlight the framework's potential to enhance urban mobility management, especially by improving parking meter placement and reducing inefficiencies and improving accessibility. Validation on real-world data from the city of Caserta, confirms the robustness and adaptivity of the proposed framework, although expanding the dataset and refining specific components are necessary for fully realizing its potential and ensuring sustainable urban planning.