車輛在高速公路之移動雲端運算系統中基於最佳化技術之卸載與資源分配策略

Abstract

自動駕駛車輛是一種新興技術，它有許多不同類型任務 (task) 的需求，包括低延遲的運算任務和資源密集型的運算任務。由於車輛的計算能力和儲存容量有限，如何服務如此大量的任務已成為車用行動通訊網路中的嚴峻挑戰。因此，本研究將使用雲端運算系統來克服車輛自身資源有限的問題，並利用高速公路有固定路線的優點，得到關於車輛移動方向和速度更佳的預測結果。 本論文主要研究在高速公路上的雲端環境中使用資源配置策略和卸載策略來對於車輛上的任務提供更好的服務。我們將此問題設計成一個數學模型，目標為最大化雲端服務提供商的收益。並以拉格蘭日鬆弛法和次梯度法為基礎的演算法來解決此問題。我們也設計一系列的實驗以測試上述演算法的表現，實驗結果顯示此演算法在多種網路情境下均能有較佳及較穩定的可行解。

Self-driving vehicle is an emerging technology which request many different types of tasks, including low-latency computation tasks and resource intensive computation tasks. Due to the limited computational capabilities and storage capacity of vehicles, serving such a large number of tasks has become a serious challenge in the vehicular network. Therefore, this study will use mobile cloud computing systems to overcome the problem of the limited resources in vehicles. By taking the advantages of the fixed route of highway, the direction and the speed of vehicles can be more predictable. In this thesis, we focus on using resource allocation strategy and offloading strategy better serve vehicle tasks in the cloud environment on the highway. We formulate the problem as a linear integer programming problem, in which the objective is to maximize the revenue of the cloud service provider. An algorithm based on the Lagrangian relaxation method and the subgradient method is used to solve this problem. A series of experiments are designed to test the performance of the algorithm. The experimental results show that the algorithm can have better and more stable feasible solutions under various network scenarios.