From Small-scale Data Dissemination to Large Data Transfers in Vehicular Environments
Vehicular networks are an active area for research and development that is becoming an integral part of the communication infrastructure of tomorrow. A wide range of services are emerging from the rapid technological advances in vehicle's sensing, computing, and wireless capabilities. In the first part of this thesis we focus on vehicle to vehicle (V2V) communication developed for a certain set of services meant to warn or inform the occupants of a vehicle. We identify several major components representing a set of `behaviors', common in dissemination protocols in VANETs and we classify them according to their main functionalities. We distinguish three major components, namely: (i) beaconing, (ii) broadcast containment and reliability, and (iii) delay tolerance. Through extensive simulations we evaluate each individual component according to a particular set of performance metrics in specific scenarios conceived to trigger the desired responses from each module. Our results provide clearer insights into the mechanics of a dissemination protocol and show which approach is applicable in a given set of environmental parameters. Based on the observations, we propose Servus, a broadcast-based dissemination protocol that efficiently tackles the challenges of the environment. Servus is fully distributed and only requires local information. It is designed to ensure reliability, overcome transmission gaps, and handle intersections. Its performance is evaluated through simulations of both city and highway scenarios. The results show that Servus is highly efficient in limiting the number of retransmissions, while ensuring propagation over transmission gaps and maintaining a high-delivery ratio despite the environmental constraints. After showing how the challenges of inter-vehicular communication can be overcome, we turn our attention to the role of the vehicle in future communication architectures. We argue in favor of a vehicle that acts as a data-packet, piggybacking large amounts of information while traveling over the existing road infrastructure. Thus, we investigate the possibility of leveraging vehicular mobility to transport large amounts of delay-tolerant data between geographically distinct points of interest.
To this end, we propose to take advantage of the existing worldwide road infrastructure as an offloading channel to help the legacy Internet assuage its burden. This creates an alternative solution with a network capacity in the petabyte range. Using a publicly available data-set, we calculate the theoretical capacity of the proposed solution and we study the influence of several environmental parameters on the data delivery latency. We show how even in low traffic conditions, our system is able to deliver 1PB of data in less then nine hours. Then, to allow for a more thorough investigation of its potential, we extend our study with tests on a broader selection of highways comprising a large variety of environmental parameters. As other solutions for massive data transfers already exist, we compare our system with an internet-based approach and a package delivery solution. The results show that our proposal outclasses the internet-based approach in terms of throughput, obtaining a performance increase of up to 260 times. The package-delivery solution achieves delivery delays that are at least twice as large as ours. We finally investigate an incentive-based approach meant to increase the penetration ratio of data-carrying vehicles. We evaluate the costs of such an approach and compare it to other alternatives. Our analysis shows that it would be extremely inefficient to lease dedicated lines in order to achieve the transport capacity of a vehicular-based transfer system. Our work will help to better understand the impact of different requirements and environmental constraints on existing and emerging applications. By properly exploiting the characteristics of the vehicular environment we encourage the development of new research topics exploiting the functionalities of the vehicles of tomorrow.
Defence : 07/16/2013 - 14h30 - Site Jussieu 25-26/105 Jury members : Yacine Ghamri-Doudane (ENSIIE) [Rapporteur]
Nathalie Mitton (INRIA) [Rapporteur]
Giovanni Pau (UPMC)
Bertrand Ducourthial (Université de Technologie de Compiégne)
Prométhée Spathis (UPMC)
Serge Fdida (UPMC)
P. Spathis, R. Gorcitz, R. Wakikawa, S. Fdida : “Dissecting Dissemination in VANETs”, AINTEC'12 - ACM Asian Internet Engineering Conference, Bangkok, Thailand, pp. 62-68, (ACM) (2012)
R. Gorcitz, Y. Jarma, P. Spathis, M. Dias de Amorim, R. Wakikawa, J. Whitbeck, V. Conan, S. Fdida : “Vehicular Carriers for Big Data Transfers”, 2012 IEEE Vehicular Networking Conference (VNC), Seoul, South Korea, pp. 109-114, (IEEE) (2012)