Traffic Simulation with Incomplete Data: the Case of Brussels

Davide Andrea Guastella, Bruno Cornelis, Gianluca Bontempi

Abstract: Intelligent transport systems are intended to support the decision-making process for reducing traffic congestion, traffic accidents, and air pollution, these having an impact on citizens' well-being. Decision-making tasks are complex because of the unpredictable dynamics of urban traffic. To tackle this complexity, simulation tools enable evaluating in silico the impact of policies on urban infrastructures. Accurate and continuous information about traffic is necessary to define simulation models that reflect the dynamics of the real traffic. However, this is not always possible, as data from the physical environment are collected generally by sensors that undergo maintenance or unpredictable temporary failures, resulting in sparse data sets that cannot be used to create accurate simulation models.We propose an approach for going from sparse data to traffic simulation models. We use the HybridIoT technique to estimate missing traffic data, used to create traffic simulation models, and SUMO, an open-source traffic simulation tool, to simulate traffic. We integrate data provided from two traffic services that operate in the city of Brussels (Belgium). We also simulate the lack of different percentages of data (from 10% to 90%), and evaluate the accuracy of traffic models created using the estimated data.The outcome of this study is twofold: (i) the definition of a novel traffic simulation model for the city of Brussels by integrating sparse data sets, and (ii) the evaluation of the impact of missing data in the accuracy of traffic simulation models.