| Summary: | Smart mobility initiatives encompass innovative methods to support traffic management experts in decisions for how to improve urban infrastructures and reduce carbon footprint. Accurate and continuous information about traffic is necessary to implement effectively such decisions. This is not always possible because of the cost of the information: it is not possible to install sensor devices at large scale because of financial costs and privacy; employing a plethora of sensors requires significant computational capabilities to process the generated data. A centralized data analysis can hinder real-time applications, and limit their practical deployment in traffic management systems. This paper introduces a novel privacy-aware method for estimating traffic density using edge computing and without over-deploying privacy-intrusive surveillance technologies such as cameras. The objective is to reduce the cost of collecting data while providing accurate information to support traffic operators in decision making. We evaluate the proposed solution using a realistic traffic data of Bologna in Italy. Results shows that it yields a 45% lower average estimation error compared to standard prediction methods. Virtual traffic monitoring devices are associated with software agents that collect data from simulated traffic and estimate traffic density measurements when this information is not available. In our experiments, when we replace 50% of camera devices with cooperative low-cost edge devices, we obtain an average percentage error of just 22%. This result indicates that the cooperation between virtual traffic monitoring devices offers a means to avoid massive deployment of camera surveillance devices using low-cost information provided by connected vehicles. We also compared the results to those obtained by standard regression techniques.
|
|---|