面向大规模动态异构网络的支持多用户并发任务的物联网应用构建方法研究

With the increase of the number of existing network resources in Internet of Things, building user application by compositing existing equipment will become the mainstream mode of the IoT application development. However, the application of the personalized demand increases the difficulty and cost of constructing IoT application. Strong heterogeneity and high dynamic of the network increase the difficulty to keep the application running robustness. Massive application number makes the resource constrained devices need to support a large amount of concurrent tasks. This project aims to put forward the method of constructing a IoT application, making the same application can run in a variety of different network environment, promoting the applicability of the application, avoiding the high cost of developing applications because of personalized requirement; This project also aims to propose a run-time resource active instantiation mechanism, through fuzzy choice to reduce the effects of heterogeneous networks, and reduce network node failure probability of the initiative to complete the selection and mapping based on resources, ensuring the robustness of application execution; This project also aims to put forward a matter similarity redundancy optimization overhead mechanism between the networking applications, improving the execution efficiency of concurrent tasks in resource constrained devices. The project is aim to explore the key issues of future mass users to share the network infrastructure and its solutions, which has important theory value and practical significance.

随着接入Internet的物联网设备数目的增多，组合物联网中已有资源来构建用户应用将成为未来物联网应用开发的主流模式。然而，应用需求的个性化增加了物联网应用构建时的难度与成本，网络的强异构性与高动态性提高了保持应用运行时鲁棒性的难度，应用数目的海量性使得资源受限设备需要支持较大的并发任务量。本项目提出一种应用与执行分离的物联网水平化应用构建方法，使得同一个应用程序可运行在多种不同的网络环境中，提升应用程序的适用性，避免由于应用个性化导致的应用开发成本过高的问题；提出一种运行时资源主动实例化机制，通过资源模糊选择降低网络异构性的影响，同时基于网络节点的失效概率主动完成资源选择与映射，保证应用程序执行的鲁棒性；提出一种利用物联网应用之间的相似性优化冗余开销的机制，提高资源受限设备的并发任务执行效率。本项目探索未来海量用户共同使用物联网基础设施时将面临的关键问题及其解决方案，具有重要的理论价值。

由于实体服务的异构性、大规模性、动态可用性、情境感知性、时空相关性、资源受限性和执行冲突性，构建通用大规模物联网系统面临挑战。本项目对物联网体系结构的若干方面展开了深入研究：在物联网平台建设方面，针对现有方法开销大、耗时长的缺点，我们提出一种分布式自适应物联网设备注册方法，将复杂且耗时的周期探测过程分配给多个设备执行，在设备动态性较高时能降低67%左右的开销，同时能使云平台降低36%左右的突发性开销，并以智能家居为例提出了一套基于规则的物联网应用构建方法；在物联网数据分析与预测方面，提出了一种基于智能手机的出行方式识别方法EasiTMC，用以支持基于用户情景和上下文的应用执行优化框架，同时提出了一种基于出行方式及语义轨迹的位置预测模型；在物联网测试系统方面，提出了一种可配置的仿真数据生成方法，以少量真实数据为输入，通过数据回归策略生成场景中任意地点与时刻传感数据的方法，同时针对物联网测试中的海量资源生成问题，提出了一套基于少量真实数据以较高的精度模拟物联网前端设备行为特征的物联网海量测试数据生成方法。