【作者】 许建龙；

【导师】 刘桂雄；

【作者基本信息】 华南理工大学 ， 制造工程智能化检测及仪器， 2013， 博士

【摘要】 物联网感知层处于物联网体系架构的基础底层，占据着重要地位。论文研究物联网平台感知层建模与快速调度方法研究及应用，对促进制造信息化技术、网络化智能测控技术的发展与应用，具有重要的学术价值和实际意义。研究得到广东省高层次人才项目(粤教师函[2010]79号)、广东省科技攻关项目(2009B010900045)、2012年粤港关键领域重点项目(2012A090200005)资助。论文首先从混合逻辑系统MLD建模、求解调度问题算法、调度死锁问题等综述论文相关内容的国内外研究进展，确定论文研究内容。论文首先研究物联网平台感知层MLD建模方法，利用物联网感知层感测节点、受控节点、协调器节点三类基本节点对系统应用场景进行描述，采用自动机对节点内部信息流运行机制进行抽象，实现物联网感知层在多类终端、多重任务、多种模式下物联网感知层整体结构体系的清晰表达，及感知层结构和动态行为的形象化、规范化描述；采用MLD建模方法，将物联网感知层整个测控行为在信息获取、调度决策、执行决策三状态中任两个状态之间自动演变模型转换为MLD模型，推出物联网感知层测控过程行为表达式，把物联网感知层测控过程行为用包含状态变量、输入变量和辅助变量来描述，将物联网感知层测控过程行为中连续动态特性、逻辑规则及操作约束集成为带有混合整数不等式约束的状态方程，可在宏观上把握系统过程行为，考虑系统操作约束、定性知识等因素，为测控系统协调优化奠定重要基础。在MLD建模基础上，论文研究物联网感知层信息快速调度规划与性能优化策略。通过基于物联网感知层MLD模型与自动机调度模型转换分析，实现整个物联网感知层MLD模型与自动机调度模型转换；研究基于层次化自动机物联网感知层快速调度机制，研究机制中的全局任务调度与控制自动机模型、局部调度自动机模型，对不同层级系统，均可使用独立调度策略，保证感知层系统既有序、可靠又快速运行。借助UPPAAL仿真工具，结果表明，VOSM系统按照经典串行工作机制未采用调度器则性能指标较差并可能发生死锁，若采用层次化自动机调度策略，性能指标可得到明显提升。此外还研究基于死锁问题的感知层性能优化方法，推导感知层发生死锁时任务与资源分配情况数学特征表达方法，建立包括全局死锁、数据获取局部死锁、参数运算局部死锁等多个检测器的时间约束及数学表达式，应用遗传算法GA解决感知层中n任务陷入死锁快速解除死锁及保证解锁代价最小等问题。考虑到感知层中任务调度动态适应性问题，论文研究物联网感知层信息动态调度策略。提出数据优先下感知层信息动态调度方法(DP-PIDSM)实现构架，应用基于DPQ-EDF的信息服务分类方法，实现传感数据按其优先级紧急程度先后将传感数据发送到协调器。提出数据优先下基于PP-WRR的信息服务调度方法，使高优先级队列得到高概率服务，低优先级队列服务得到保障。开展能耗约束下感知层节点信息调度方法(EL-PSISM)研究，针对节点在满足更新率下节能问题，采用决策系数Cs_i对节点休眠进行调度，并用MLD方法构建唤醒状态传感节点信息智能调度模型。基于OPNET的DP-PIDSM、EL-PIDSM仿真平台，分别对30、48个传感节点并行调度，结果表明基于DPQ-EDF信息服务分类与PP-WRR信息服务调度，可有效提高性能指标；相对于SP算法、WRR算法，PP-WRR算法在包调度处理平均延迟相差0.03s、0.01s情况下，截止期满足指标分别提高了23.2%、7.3%；EL-PIDSM休眠调度与信息通信调度策略，与未采取EL-PIDSM相比，在信息损失系数为0.148时可节能50.4%，体现EL-PIDSM模型应用的可行性、有效性。为了检验提出方法的应用效果，论文论述物联网感知层调度方法在机动车运行安全状态监测(VOSM)平台的应用，从基于物联网的VOSM系统构架入手，研究系统信息调度优化模型，建立各调度模块时间约束并讨论分析其物理意义，在VOSM平台上开展快速调度、数据优先实验表明了基于物联网感知层快速调度技术的实用性、有效性。更多还原

【Abstract】 Internet of things (IoT) perception layer is the bottom layer in IoT architecture, itoccupies important position. This dissertation researches on IoT perception layer behaviormodeling method and fast scheduling method and application, which has important academicvalue and practical significance for promoting manufacture information technology,networked intelligent measurement and control technology development and application. Theresearch work is supported by High-level Talents Project of Guangdong higher school (letter[2010] No.79), Industrial Research Project of Guangdong Science and TechnologyDepartment (2009B010900045), and2012Guangdong-Hong Kong Key Area Major Project(2012A090200005).The research progress of related content at home and abroad includingmixed logical system MLD modeling, scheduling algorithm research progress, deadlockproblem with scheduling, etc. is reviewed to decide research contents of this paper. Researchwork includes several aspects as follows:This paper first studys IoT perception platform layer behavior MLD modeling method.Three basic nodes including sensor node, controlled node, coordinating node are used todescribe system application scenario, nodes’ internal information flow mechanism areabstracted by automata.. It realizes that the whole structure of IoT perception layer isexpressed clearly under multi-terminal, multi-tasking, multi-mode, the layer structure anddynamic behavior are described visually and normalized. MLD modeling method is used withIoT perception layer whole measurement and control behavior in information acquisition,scheduling, execution decisions between two states in three state automatically evolutionmodel is converted to MLD model, and IoT perception measurement and control processbehavior expression is deduced. IoT perception layer measurement and control processbehavior is described by state variables and input variables and auxiliary variables, thecontinuous dynamic feature, logical rules and operation constraints are integrated into stateequation with mixed integer inequality constraints.It can can grasp system process, systemoperation constraints, the qualitative knowledge and other factors on a macro level, it lays aimportant foundation for measurement and control system coordination optimization.Based on the MLD modeling, this paper studys perception layer scheduling planning method and performance optimization strategy. Based on IoT perception layer MLD modeland automaton scheduling model transformation analysis, it can realize the whole IoTperception layer MLD model and automatic scheduling model transformation. IoT perceptionlayer fast scheduling mechanism is proposed based on hierarchical automata. Global taskscheduling and control automata model, local scheduling automaton model are studied in themechanism. Systems at different levels can use independent scheduling policy and realize thatthe perception layer runs orderly, reliably and quickly. VOSM system hierarchical automatascheduling strategy is studied by UPPAAL simulation tool, results show that the performanceof VOSM system according to classic serial working mechanism not using the scheduler ispoor performance indicators and deadlocks may occur,while using hierarchical automatascheduling strategy, the performance index improved significantly. Furthermore, perceptionlayer performance optimization method is studied based on the deadlock problem. Themathematics expression is deduced for task and resource distribution characteristics when adeadlock occurs in perception layer. Time constraints and mathematical expression areestablished for deadlock detectors including global deadlock, data acquisition deadlock,parameter computing local deadlock. Genetic algorithm (GA) is used to quickly reliefdeadlock and ensure the cost is minimum when multitask stuck in a deadlock in perceptionlayer.According to the dynamic adaptability of task scheduling problem in perception layer, thispaper studys IoT perception layer information dynamic scheduling strategy. Implementationarchitecture of perception layer information dynamic scheduling method under data priority(DP-PIDSM) is put forward. Information service classification method based on theDPQ-EDF is adopted to realize sensor data sent to the coordinator according to the priority ofthe emergency degree to sensor data. Information scheduling method based on PP-WRR isput forward, it guarantee high priority queues can be served in high probability and lowpriority queues are served in low probability. Perception layer dynamic scheduling methodunder energy limit (EL-PSISM) is put forward. According to problems of how to save energyand meet the sensor nodes update rate, decision coefficient Cs_iis used to schedule nodessleep or not. Intelligent sensor node information scheduling model is built by MLD method,Simulation based on OPNET for DP-PIDSM and EL-PIDSM，parallel scheduling for30and 48sensor nodes respectively, the results shows that, DPQ-EDF information serviceclassification and PP-WRR information dynamic scheduling methods relative to the SPalgorithm, WRR algorithm, when packet scheduling process average delay is0.03s,0.01s, thedeadline index is increased by23.2%,7.3%respectively. Sleep scheduling and informationcommunication scheduling strategy, compared with not take EL-PIDSM scheduling methods,it can save energy50.4%when information loss coefficient is0.148. It reflects feasibility andeffectiveness of the model EL-PIDSM.To test the application effect of the method, the paper discusses the application of IoTperception layer scheduling method in vehicle operation safe state motoring (VOSM)platform. According to VOSM system architecture based on IoT, the system informationscheduling optimization model is researched and the time constraints for scheduling moduleare set up and their physical meaning is discussed. VOSM system fast scheduling and datapriority experiments show the practicability, effectiveness with IoT fast schedulingtechnology.更多还原