面向认知物联网的自主认知与智慧决策机制研究

Internet of Things (IOT) is a heterogeneous, mixed and uncertain ubiquitous network, the application of which is booming in the field of modern intelligent service, such as ecological protection, energy conservation & emission reduction, food security, etc. The autonomic cognition and intelligent decision-making mechanism is the core to achieve intelligence for IOT. The project focuses on three aspects of comprehensive research for IOT, including autonomic cognitive system structure, autonomic cognitive decision-making mechanism and self-learning optimization mechanism. Firstly, the project will propose the three layer cognitive ring and interaction sharing mechanism of perceptive information, which form the autonomic cognitive system structure, so as to provide infrastructure and data support for the cognitive decision-making. Secondly, the project will propose the cross-layer decision of the cognitive element and the multi-autonomous-domain cooperative decision-making model, which construct the autonomic cognitive decision-making mechanism to realize the intelligent decision-making of IOT facing QoS and Network Performance Objective (NPO). Thirdly, the project will propose the incremental learning optimization method oriented cognitive element and multi-autonomous-domain cooperative decision-making, which compose the cognitive decision-making learning optimization mechanism to constantly improve the intelligence of cognitive decision-making. The results of the research will endow the autonomic and intelligent features for IOT, forming the cognitive IOT and supporting it to realize the efficient intelligent decision-making and improve network performance, which will have wide application prospect and great scientific significance.

物联网是一种异构、混杂、不确定的泛在网络，在生态保护、节能减排、食品安全等现代智慧服务领域应用发展迅速，自主认知与智慧决策机制是体现物联网智慧特性的关键。本项目重点围绕物联网的自主认知体系结构、自主认知决策机制和学习优化机制三方面展开综合研究：①提出三层认知环和感知信息交互共享机制，构成自主认知体系结构，从而为认知决策提供基础架构和数据支撑；②提出认知元跨层决策和多自治域协作决策模型，构成自主认知决策机制，从而实现面向QoS和网络性能目标的智慧决策；③提出面向认知元决策和多域协作决策的增量式学习优化方法，构成认知决策学习优化机制，从而不断提高认知决策的智慧。研究成果赋予物联网自主、智慧特性，形成认知物联网，并支持认知物联网实现高效智慧决策和提高网络性能的目的，具有广泛的应用前景和重大的科学意义。

作为一种急速发展的超大规模网络，物联网在生态保护、节能减排、智能电网、食品安全等现代智慧服务等领域应用广泛。物联网组成要件的复杂性、应用形态的不确定性和运行规律的模糊性使得智慧特性成为决定其自身和拓展领域发展的关键属性，其异构性为跨域、跨子网的智慧性研究带来了巨大的挑战。在此背景下，本项目重点围绕物联网的自主认知体系结构、自主认知决策机制和学习优化机制三方面展开综合研究：(1)在物联网现有功能的基础上，引入认知元素，增加自感知、自决策、自学习、自优化、自调节等智慧特性，提出了认知物联网三维体系结构、三层认知环结构及相应的协议栈模型，构成自主认知体系结构；在自主认知体系结构的基础上，提出了一种不完备感知信息的填补算法，提高感知信息的可用率；提出了基于网络感知的路由机制与方法，实现域内感知信息的交互，为认知决策提供了基础架构和数据支撑。(2)面向认知物联网自治域内的QoS要求和性能目标，基于获取的感知信息，分析影响物联网认知决策的诸多因素。当单个认知元无法满足用户QoS和网络性能目标的情况下，利用多个认知元的协作完成决策；当单个域无法满足用户QoS和网络性能目标的情况下，利用多域协作完成决策。此外，提出了认知元跨层决策、多自治域协作决策模型、基于跨层的域内/域外协作路由决策方法，构成自主认知决策机制，从而实现面向QoS和网络性能目标的智慧决策。(3) 智能决策和机器学习密不可分，没有学习的决策只能是静态决策、固定决策，只有在决策过程中引入学习机制，才能在决策过程中不断学习，逐步提高决策的智慧，才是真正的智慧决策。据此，提出了物联网安全态势自律感知与评估、安全协作优化及学习等，构成认知决策学习优化机制，从而不断提高认知决策的智慧。项目研究发表SCI收录期刊论文5篇、EI收录期刊论文8篇、EI/CPCI收录国际会议论文4篇，出版专著2部，申请发明专利5项，获河南省科技进步三等奖1项，培育中青年学术带头人3人，硕士研究生6人。项目研究成果赋予物联网自主、智慧特性，支持认知物联网实现高效智慧决策和提高网络性能的目的，具有广泛的应用前景和重大的科学意义。