Mobility Trace Analysis in the Design of Vehicular Networks

Abstract

Vehicular networks have emerged as a promising technology for efficient data communication in transportation systems and smart cities. At the same time, the popularization of devices with attached sensors has allowed obtaining a large volume of data with spatiotemporal information from different entities. In this sense, we face a significant amount of vehicular mobility data (mobility traces) being recorded. Those traces provide unprecedented opportunities to understand the dynamics of vehicular mobility and provide data-driven solutions. For instance, we can observe the nature of the network topology or propose routing protocols by looking at the historical movements of vehicles in a city.

In this work, we delve into the practical applications of mobility traces in the design of vehicular networks. We start by identifying the key characteristics of publicly available vehicular mobility traces, using a list of criteria and performing a characterization. We also address the issue of inconsistencies in traces, proposing two solutions that not only repair these inconsistencies but also demonstrate their impact on the design of vehicular networks. Additionally, we propose two frameworks for generating high-quality bus mobility traces, which can be invaluable in simulating vehicular networks and urban mobility studies.

Moreover, we aim to identify the peculiarities of the vehicular network topology obtained from real-world mobility traces. To this end, we perform several analyses showing the strengths and weaknesses of these approaches that consider a dynamic network standpoint. We reveal and model how mobility dynamics impact vehicular networks composed of buses and the formation of vehicular clouds.

Last but not least, we develop routing protocols that consider characteristics extracted from vehicle mobility in the message-routing process on the network. Through simulations based on realistic mobility traces, we show the applicability and viability of these protocols in terms of latency, delivery rate, and overhead.