以失真為基礎的車際網路影像串流傳輸機制

Abstract

雖然至今為止已有許多針對VANET特性及需求而提出的路由協定，但是大部分的這些研究工作並不適合用來傳遞即時影像串流。因為VANET非常容易因為車輛的行駛行為而造成網路局部性的斷線以及整個網路拓樸的改變。這些與生俱來的天性使得許多現有的路由協定很難能夠滿足影像串流或是其它多媒體相關的應用對於封包延遲時間的要求。為了解決這樣的問題，本論文提出了一個基於交通統計資訊以及影像失真程度的路由協定。這個協定主要是透過將每條道路路段指定一個估計的失真數值以作為該路段是否適合用來傳遞影像封包的權重，然後我們可以據此找出最適合傳影像的路段組合。同時，本論文也提出一個應用Markov Decision Process的概念的封包傳遞機制，來幫助持有影像封包的車輛找出下一個適合的封包轉傳車輛。在此我們採用NS2網路模擬器來評估我們演算法的效能，模擬結果顯示出我們的方法的確能讓使用者獲得更佳的視訊品質。

There are already many routing protocols tailored for Vehicular Ad Hoc Networks (VANET). However, most of them are not suitable for real-time video streaming. The frequently isolating network partitioning nature and constantly changing network topology of VANET makes it difficult for existing protocols to satisfy the strict delay requirements of video streaming or other multimedia applications. To address this issue, this thesis proposes a routing protocol which considers both the traffic statistics information and possible video distortion to assign different weights to each road segment, and find the best path for video packet dissemination accordingly. This thesis also proposes a Markov Decision Process (MDP) based forwarding scheme for the process of packet delivery. The simulations are carried out by NS2 to validate the performance of our proposed protocol. The results show that the proposed solution can really yield good user perceived video quality.