軟體定義車載網路之延遲與成本最佳化

Abstract

車載隨意網路(Vehicular ad hoc network；VANET)是一種透過移動中的車輛及路邊設施為節點所形成的行動網路。隨著相關研究與技術的進展，有人提出了應用軟體定義網路(software-defined network；SDN)於車載網路之中。SDN的特性提高了VANET的可編程性，於是產生了軟體定義車載網路(software-defined vehicular network；SDVN)。然而，SDVN確仍然有些不足之處，如車輛異質性的考量、頻寬資源有限、路徑的不確定性、延遲時間過長等等。近年來雖然已有許多研究試圖解決這些問題且有良好的成效，但延遲時間過長的問題缺乏一個妥善的解決方案。在SDVN的環境中，解決延遲時間的問題就是使用快速的蜂巢狀線路(cellular link)進行通訊，但這帶來的負面效果便是成本高。要解決此問題，需要一個有系統性的策略規劃。因此，本文將針對延遲和成本的問題進行探討，提出一個回饋獎勵的策略模型，透過回扣率鼓勵車輛駕駛使用蜂巢狀線路。為求找到一個兩者皆能達到最優的方式，使成本和延遲盡可能的降低。並非類似部分文獻兩階段的處理問題，而是利用同時求解的方式，找出兩者的最佳解。並透過本研究提出改良型的基因演算法(improved genetic algorithm；IGA)來解決問題，IGA是使用基因演算法為基礎，整合了突變率的動態調整機制，保證了求解的多樣性，避免快速收斂導致陷入局部最佳解。在此同時也設計了保護機制，保證最佳染色體並不會因為突變的緣故而被破壞。實驗結果顯示使用蜂巢狀線路的量會隨著回扣率提高，使用蜂巢狀線路的量也會與其他參數呈現正相關。此外透過提出的IGA的方法能顯著提升求解的效果，而增加使用者效益函數得到的結果也比過去的好。

Vehicular ad hoc network (VANET) is a mobile network technology formed by moving vehicles and roadside units. Due to the progress of related research and technology, some people apply software-defined network (SDN) in VANET. SDN features effectively improve the programmability of VANET, and there is a software-define vehicular network (SDVN). However, SDVN still has some problems (e.g., heterogeneous vehicles, limited bandwidth, trajectory uncertainty, latency). Therefore, many studies have tried to solve these problems and have good results. However, the problem of latency control lacks a proper solution. In the SDVN environment, the problem of settling the delay time is to use the slightly faster cellular link with communication, but the effect of this method is the high cost. To solve this problem, we need a systematic strategy. Therefore, this paper will focus on the issue of delay and cost, and propose a rebating bandwith strategy model. In order to find a way both can achieve the best way to make the cost and delay as lower as possible. Is not similar to some of the two-stage processing problems, but the use of simultaneous solution to find the best solution for both. We will solve the problem through the improved genetic algorithm (IGA). IGA is based on genetic algorithm (GA), which ensures the solution diversity by integrates the dynamic adjustment mechanism of mutation rate, to avoid rapid convergence lead to fall into the local best solution. At the same time, a protection mechanism was designed to ensure that the best chromosomes were not destroyed by the dynamic mutation. The experimental results show that the amount of the use of the cellular line will increase with the rebate rate, and the amount of the cellular line will be positively correlated with the other parameters. In addition, the proposed IGA method can significantly improve the effect of solving, and increase the user benefit function to get the results better than in the past.