Real-time video transmission in multi-homed vehicular networks

Providing high quality video transmission in vehicular ad-hoc networks (VANETs) is very challenging due to the highly dynamic, unpredictable topology, and low bandwidth characteristics. Moreover, within an IP-based multihomed VANET, there is an array of different network access technologies roadside...

ver descrição completa

Detalhes bibliográficos
Autor principal: Lopes, Rui Eduardo de Figueiredo Arnay (author)
Formato: masterThesis
Idioma:eng
Publicado em: 2019
Assuntos:
Texto completo:http://hdl.handle.net/10773/29561
País:Portugal
Oai:oai:ria.ua.pt:10773/29561
Descrição
Resumo:Providing high quality video transmission in vehicular ad-hoc networks (VANETs) is very challenging due to the highly dynamic, unpredictable topology, and low bandwidth characteristics. Moreover, within an IP-based multihomed VANET, there is an array of different network access technologies roadside units (RSUs), through which on-board units (OBUs) establish connections. In these connections, each available technology has its own medium access control protocol, and other defined parameters, that may interfere with the packet’s order of arrival in the mobile network’s clients, attached to an OBU. In this document, we design a system able to optimally transmit RTP video streams in such IP-based multi-homed VANETs. By splitting the video stream packet flows through its different frame types, in the array of available multihomed paths, the system can then send critical frames, depending on the used coding standard, through a more reliable path. Tests were conducted with H.264 encoded video streams, differing in spatial resolutions of 240p, 360p, 720p, and 1080p, between three different video frame distribution algorithms. Such tests were also executed in three different phases (on both laboratory and outside environments). This dissertation results show that without applying any criterium to video streams distribution, in the VANET’s manager, the video transmission loss percentages are significant, mostly in smaller resolutions. Otherwise, with the video frame distribution algorithms enabled, we achieved decreases until approximately 60.4%, as evaluated in our outside experiments.