Studies on Design and Evaluation of a Video Multicast Method for Heterogeneous Mobile Terminals

在这项研究中，我们已经考虑了一个具有成本效益的视频流到异构移动的终端通过内容分发网络（CDN）的组播方法。基于这样的要求，使得每个移动的终端可以在显示器尺寸、计算功率、可用电池功率和带宽的约束下指定接收视频的质量，并且每个终端可以接收最满足用户需求的视频，我们具有以下结果。(1)原始视频转换成各种质量，让代理执行转码服务的基础上用户的要求。我们已经开发了一种算法，用于决定服务器，代理和移动的终端之间的连接路径，以最大限度地减少CDN的资源。此外，我们还设计了一个协议，用于根据用户的需求周期性地重建路径。(2)我们考虑了一种在移动的ad hoc网络中构造组播树的方法。在一般情况下，它是NP-难的多个约束条件下，如延迟时间和带宽，构建成本最小的组播树，因此，我们已经开发了一种有效的算法，使用遗传算法获得一个半最优解。为了适用于大规模的网络，我们设计了一个分布式算法，分为两个阶段：通过聚集附近的节点，将移动的节点集划分为簇。在此基础上，对每个簇构造一棵半最优组播树，然后对簇的代表节点集构造一棵半最优组播树，并通过仿真验证了上述结果的有效性。

In this research, we have considered a cost-efficient multicast method for video streaming to heterogeneous mobile terminals via a content delivery network (CDN). Based on the requirements such that each mobile terminal can specify the quality of receiving video under the constraints of display size, computation power, available battery power, and bandwidth and that each terminal can receive the video mostly satisfying the user's demand, we have the following results.(1) The original video is converted in various qualities by letting proxies execute transcoding services based on the users' requirements. We have developed an algorithm for deciding connecting paths among a server, proxies, and mobile terminals so as to minimize the resource of CDN. Furthermore, we have designed a protocol for reconstructing the paths periodically according to users' requirements.(2) We consider a method for constructing a multicast tree in a mobile ad hoc network. In general, it is NP-hard to construct the cost-minimum multicast tree under multiple constraints such as latency time and bandwidth, and thus we have developed an efficient algorithm for obtaining a semi-optimal solution using a genetic algorithm. In order to apply to a large scale network, we have designed an distributed algorithm with two stages, as follows: The set of mobile nodes is divided into clusters by gathering the nodes located near. After that, a semi-optimal multicast tree is constructed for each cluster, and then a semi-optimal multicast tree is constructed for the set of representative nodes of clusters.We have confirmed the usefulness of the results above through simulations.